norman00argument Argument in Multi-Agent Systems Multi-agent systems research is concerned both with the modelling of human and animal societies and with the development of principles for the design of practical distributed information management systems. This position paper will, rather than examine the various dierences in perspective within this area of research, discuss issues of communication and commitment that are of interest to multi-agent systems research in general. 1 Introduction A computational society is a collection of autonomous agents that are loosely dependent upon each other. The intentional stance [12] is often taken in describing the state of these agents. An agent may have beliefs, desires, intentions, and it may adopt a role or have relationships with others. Thus, multi-agent systems (MAS) as with most AI research is signi cantly inuenced, at least in its vocabulary, by philosophy and cognitive psychology. 1 So, what's the point? Computational societies are developed for two primary reasons: Mode... Agents kusiak00decomposition Decomposition in Data Mining: An Industrial Case Study Data mining offers tools for discovery of relationships, patterns, and knowledge in large databases. The knowledge extraction process is computationally complex and therefore a subset of all data is normally considered for mining. In this paper, numerous methods for decomposition of data sets are discussed. Decomposition enhances the quality of knowledge extracted from large databases by simplification of the data mining task. The ideas presented are illustrated with examples and an industrial case study. In the case study reported in this paper, a data mining approach is applied to extract knowledge from a data set. The extracted knowledge is used for the prediction and prevention of manufacturing faults in wafers. ML pant02exploration Exploration versus Exploitation in Topic Driven Crawlers Topic driven crawlers are increasingly seen as a way to address the scalability limitations of universal search engines, by distributing the crawling process across users, queries, or even client computers. The context available to a topic driven crawler allows for informed decisions about how to prioritize the links to be explored, given time and bandwidth constraints. We have developed a framework and a number of methods to evaluate the performance of topic driven crawler algorithms in a fair way, under limited memory resources. Quality metrics are derived from lexical features, link analysis, and a hybrid combination of the two. In this paper we focus on the issue of how greedy a crawler should be. Given noisy quality estimates of links in a frontier, we investigate what is an appropriate balance between a crawler's need to exploit this information to focus on the most promising links, and the need to explore links that appear suboptimal but might lead to more relevant pages. We show that exploration is essential to locate the most relevant pages under a number of quality measures, in spite of a penalty in the early stage of the crawl. IR emmerich00software Software Engineering and Middleware: A Roadmap The construction of a large class of distributed systems can be simplified by leveraging middleware, which is layered between network operating systems and application components. Middleware resolves heterogeneity, and facilitates communication and coordination of distributed components. State of-the-practice middleware products enable software engineers to build systems that are distributed across a localarea network. State-of-the-art middleware research aims to push this boundary towards Internet-scale distribution, adaptive systems, middleware for dependable and wireless systems. The challenge for software engineering research is to devise notations, techniques, methods and tools for distributed system construction that systematically build and exploit the capabilities that middleware products deliver, now and in the future. DB 386938 Dynamic-Agents for Dynamic Service Provisioning We claim that a dynamic-agent infrastructure can provide a shift from static distributed computing to dynamic distributed computing, and we have developed such an infrastructure to realize such a shift. We shall show its impact on software engineering through a comparison with other distributed object-oriented systems such as CORBA and DCOM, and demonstrate its value in highly dynamic system integration and service provisioning. The infrastructure is Java-based, light-weight, and extensible. It differs from other agent platforms and client/server infrastructures in its support of dynamic behavior modification of agents. A dynamic-agent is not designed to have a fixed set of predefined functions but instead, to carry application-specific actions, which can be loaded and modified on the fly. This allows a dynamic-agent to adjust its capability for accommodating environment and requirement changes, and play different roles across multiple applications. The above features are supported b... Agents 532373 A Hybrid Mobile Robot Architecture with Integrated Planning and Control Research in the planning and control of mobile robots has received much attention in the past two decades. Two basic approaches have emerged from these research efforts: deliberative vs. reactive. These two approaches can be distinguished by their different usage of sensed data and global knowledge, speed of response, reasoning capability, and complexity of computation. Their strengths are complementary and their weaknesses can be mitigated by combining the two approaches in a hybrid architecture. This paper describes a method for goal-directed, collision-free navigation in unpredictable environments that employs a behavior-based hybrid architecture with asynchronously operating behavioral modules. It differs from existing hybrid architectures in two important ways: (1) the planning module produces a sequence of checkpoints instead of a conventional complete path, and (2) in addition to obstacle avoidance, the reactive module also performs target reaching under the control of a self-organizing neural network. The neural network is trained to perform fine, smooth motor control that moves the robot through the checkpoints. These two aspects facilitate a tight integration between high-level planning and low-level control, which permits real-time performance and easy path modification even when the robot is en route to the goal position. Agents arge02optimal Optimal External Memory Interval Management In this paper we present the external interval tree, an optimal external memory data structure for answering stabbing queries on a set of dynamically maintained intervals. The external interval tree can be used in an optimal solution to the dynamic interval management problem, which is a central problem for object-oriented and temporal databases and for constraint logic programming. Part of the structure uses a weight-balancing technique for efficient worst-case manipulation of balanced trees, which is of independent interest. The external interval tree, as well as our new balancing technique, have recently been used to develop several efficient external data structures. DB pandurangan02using Using PageRank to Characterize Web Structure Recent work on modeling the Web graph has dwelt on capturing the degree distributions observed on the Web. Pointing out that this represents a heavy reliance on "local" properties of the Web graph, we study the distribution of PageRank values (used in the Google search engine) on the Web. This distribution is of independent interest in optimizing search indices and storage. We show that PageRank values on the Web follow a power law. We then develop detailed models for the Web graph that explain this observation, and moreover remain faithful to previously studied degree distributions. We analyze these models, and compare the analyses to both snapshots from the Web and to graphs generated by simulations on the new models. To our knowledge this represents the first modeling of the Web that goes beyond fitting degree distributions on the Web. IR 18832 XM2VTSDB: The Extended M2VTS Database In this paper we describe the acquisition and content of a large multi-modal database intended for training and testing of multi-modal verification systems. The XM2VTSDB database offers synchronised video and speech data as well as image sequences allowing multiple views of the face. It consists of digital video recordings taken of 295 hundred subjects at one month intervals taken over a period of five months. We also describe a protocol for evaluating verification algorithms on the database. The database has been made available to anyone on request to the University of Surrey through http://www.ee.surrey.ac.uk/Research/VSSP/xm2vtsdb. DB binder01design Design And Implementation Of The J-Seal2 Mobile Agent Kernel J-SEAL2 is a secure, portable, and efficient execution environment for mobile agents. The core of the system is a micro-kernel fulfilling the same functions as a traditional operating system kernel: protection, communication, domain termination, and resource control. This paper describes the key concepts of the J-SEAL2 micro-kernel and how they are implemented in pure Java. Agents slivinskas00foundation A Foundation for Conventional and Temporal Query Optimization Addressing Duplicates and Ordering AbstractÐMost real-world databases contain substantial amounts of time-referenced, or temporal, data. Recent advances in temporal query languages show that such database applications may benefit substantially from built-in temporal support in the DBMS. To achieve this, temporal query representation, optimization, and processing mechanisms must be provided. This paper presents a foundation for query optimization that integrates conventional and temporal query optimization and is suitable for both conventional DBMS architectures and ones where the temporal support is obtained via a layer on top of a conventional DBMS. This foundation captures duplicates and ordering for all queries, as well as coalescing for temporal queries, thus generalizing all existing approaches known to the authors. It includes a temporally extended relational algebra to which SQL and temporal SQL queries may be mapped, six types of algebraic equivalences, concrete query transformation rules that obey different equivalences, a procedure for determining which types of transformation rules are applicable for optimizing a query, and a query plan enumeration algorithm. The presented approach partitions the work required by the database implementor to develop a provably correct query optimizer into four stages: The database implementor has to 1) specify operations formally, 2) design and prove correct appropriate transformation rules that satisfy any of the six equivalence types, 3) augment the mechanism that determines when the different types of rules are applicable to ensure that the enumeration algorithm applies the rules correctly, and 4) ensure that the mapping generates a correct initial query plan. Index TermsÐTemporal databases, query optimization, transformation rules, temporal algebra, duplicate elimination, coalescing. 1 DB 496719 Use of Satellite Image Referencing Algorithms to Characterize Asphaltic Concrete Mixtures A natural way to test the structural integrity of a pavement is to send signals with different frequencies through the pavement and compare the results with the signals passing through an ideal pavement. For this comparison, we must determine how, for the corresponding mixture, the elasticity E depends on the frequency f in the range from 0.1 to 10 5 Hz. It is very expensive to perform measurements in high frequency area (above 20 Hz). To avoid these measurements, we can use the fact that for most of these mixtures, when we change a temperature, the new dependence changes simply by scaling. Thus, instead of performing expensive measurements for different frequencies, we can measure the dependence of E on moderate frequencies f for different temperatures, and then combine the resulting curves into a single "master" curve. In this paper, we show how fuzzy techniques can help to automate this "combination". ML kitano99robocup RoboCup Rescue: Search and Rescue in Large-Scale Disasters as a Domain for Autonomous Agents Research Disaster rescue is one of the most serious social issue which involves very large numbers of hetergenious agents in the hostile environment. RoboCup-Rescue intends to promote research and development in this socially significant domain by creating a standard simulator and forum for researchers and practitioners. While the rescue domain intuitively appealing as large scale multi-agent domains, it has not yet given through analysis on its domain characteristics. In this paper, we present detailed analysis on the task domain and elucidate characteristics necessary for multi-agent systems for this domain. 1 Introduction In this paper, we propose RoboCup-Rescue, as a secondary domain for RoboCup activities [ Kitano, et al., 1997]. The aim of RoboCup-Rescue are (1) to ensure smooth transfer of technologies invented through RoboCup activity to a socially significant real world domain, (2) to establish a domain which complements features that are missing in soccer, and (3) to examine funderm... HCI leonhardi99virtual Virtual Information Towers - A Metaphor for Intuitive, Location-Aware Information Access in a Mobile Environment This paper introduces Virtual Information Towers (VITs) as a concept for presenting and accessing location-aware information with mobile clients. A VIT is a means of structuring location-aware information, which is assigned to a certain geographical position while having a certain area of visibility. A user equipped with a mobile, wearable computer has access to the VITs which are "visible" from his/her current location. The architecture and protocols of a system are described, which allows its users to create VITs and to access the information on them using Internet mechanisms. We have implemented a prototype of this system and a VIT client for a wearable computer and will present some aspects of this implementation. 1 Introduction Mobile information access is an important field of application for wearable computers. In a mobile environment much of the accessed information is location-dependent, i.e. the content of the information or the user's interest in the information depends on ... HCI prasad96offline Off-line Learning of Coordination in Functionally Structured Agents for Distributed Data Processing When we design multi-agent systems for realistic, worth-oriented environments, coordination problems they present involve intricate and sophisticated interplay between the domain and the various system components. Achieving effective coordination in such systems is a difficult problem for a number of reasons like local views of problem-solving task and uncertainty about the outcomes of interacting non-local tasks. In this paper, we present a learning algorithm that endows agents with the capability to choose an appropriate coordination algorithm based on the present problem solving situation in the domain of distributed data processing. 1 Introduction Achieving effective coordination in a multi-agent system is a difficult problem for a number of reasons. The first is that an agent's control decisions, based only on its local view of problem-solving task structures, may lead to inappropriate decisions about which activity it should do next, what results it should transmit to other agen... ML datta98case A Case for Parallelism in Data Warehousing and OLAP In recent years the database community has experienced a tremendous increase in the availability of new technologies to support efficient storage and retrieval of large volumes of data, namely data warehousing and On-Line Analytical Processing (OLAP) products. Efficient query processing is critical in such an environment, yet achieving quick response times with OLAP queries is still largely an open issue. In this paper we propose a solution approach to this problem by applying parallel processing techniques to a warehouse environment. We suggest an efficient partitioning strategy based on the relational representation of a data warehouse (i.e., star schema). Furthermore, we incorporate a particular indexing strategy, DataIndexes, to further improve query processing times and parallel resource utilization, and propose a preliminary parallel star-join strategy. 1 Introduction In recent years, there has been an explosive growth in the use of databases for decision support. This phenome... DB 100598 Error-Driven Pruning of Treebank Grammars for Base Noun Phrase Identification Finding simple, non-recursive, base noun phrases is an important subtask for many natural language processing applications. While previous empirical methods for base NP identification have been rather complex, this paper instead proposes a very simple algorithm that is tailored to the relative simplicity of the task. In particular, we present a corpus-based approach for finding base NPs by matching part-ofspeech tag sequences. The training phase of the algorithm is based on two successful techniques: first the base NP grammar is read from a "treebank" corpus; then the grammar is improved by selecting rules with high "benefit" scores. Using this simple algorithm with a naive heuristic for matching rules, we achieve surprising accuracy in an evaluation on the Penn Treebank Wall Street Journal. 1 Introduction Finding base noun phrases is a sensible first step for many natural language processing (NLP) tasks: Accurate identification of base noun phrases is arguably the most critical comp... IR snoek02stateart A State-of-the-art Review on Multimodal Video Indexing Efficient and effective handling of video documents depends on the availability of indexes. Manual indexing is unfeasible for large video collections. Effective indexing requires a multimodal approach in which either the most appropriate modality is selected or the different modalities are used in collaborative fashion. In this paper we focus on the similarities and differences between the modalities, and survey several methods aiming at automating the time and resource consuming process of video indexing. Furthermore, we put forward a unifying and multimodal framework, which views a video document from the perspective of its author. This framework forms the guiding principle for identifying index types, for which automatic methods are found in literature. It furthermore forms the basis for categorizing these different methods. IR zaiane98discovering Discovering Web Access Patterns and Trends by Applying OLAP and Data Mining Technology on Web Logs As a confluence of data mining and WWW technologies, it is now possible to perform data mining on web log records collected from the Internet web page access history. The behaviour of the web page readers is imprinted in the web server log files. Analyzing and exploring regularities in this behaviour can improve system performance, enhance the quality and delivery of Internet information services to the end user, and identify population of potential customers for electronic commerce. Thus, by observing people using collections of data, data mining can bring considerable contribution to digital library designers. In a joint effort between the TeleLearning-NCE project on Virtual University and NCE-IRIS project on data mining, we have been developing the knowledge discovery tool, WebLogMiner, for mining web server log files. This paper presents the design of the WebLogMiner, reports the current progress, and outlines the future work in this direction. DB kosch99managing Managing the Operator Ordering Problem in Parallel Databases This paper focuses on parallel query optimization. We consider the operator problem and introduce a new class of execution strategies called Linear-Oriented Bushy Trees (LBT). Compared to the related approach of the General Bushy Trees (GBT) a significant complexity reduction of the operator ordering problem can be derived theoretically and demonstrated experimentally (e.g. compared with GBTs, LBTs authorize optimization time improvement that can reach up-to 49%) without loosing quality. Finally we demonstrate that existing commercial parallel query optimizers need little extension modifications in order to handle LBTs. Key words: Parallel databases, parallel query optimization, linear-oriented bushy trees, extending existing optimizers. 1 Introduction Modern database applications, such as data mining and decision support pose several new challenges to query optimization and processing [1]. One of the main issues concerns the processing of complex queries (e.g. recent Teradata rela... DB vanthong00speechbot SpeechBot: a Speech Recognition based Audio Indexing System for the Web We have developed an audio search engine incorporating speech recognition technology. This allows indexing of spoken documents from the World Wide Web when no transcription is available. This site indexes several talk and news radio shows covering a wide range of topics and speaking styles from a selection of public Web sites with multimedia archives. Our Web site is similar in spirit to normal Web search sites; it contains an index, not the actual multimedia content. The audio from these shows suffers in acoustic quality due to bandwidth limitations, coding, compression, and poor acoustic conditions. The shows are typically sampled at 8 kHz and transmitted, RealAudio compressed, at 6.5 kbps. Our word-error rate results using appropriately trained acoustic models show remarkable resilience to the high compression, though many factors combine to increase the average word-error rates over standard broadcast news benchmarks. We show that, even if the transcription is inaccurate, we can st... IR 362156 Context Awareness by Analysing Accelerometer Data In this paper we describe continuing work being carried out as part of the Bristol Wearable Computing Initiative. We are researching processing techniques for data from accelerometers which enable the wearable computer to determine the user's activity. We have experimented with, and review, techniques already employed by others; and then propose new methods for analysing the data delivered by these devices. We try to minimise the number of devices needed, and use a single X-Y accelerometer device. Using our techniques we have adapted our GPS based Tourist Guide wearable Computer application to include a multimedia presentation which gives the user information using different media depending on the user's activity as well as location. 1 Introduction and Background This is a condensed version of a technical report. [1] Our interests in wearable computing are centred around determining the context of the user and developing applications which make use of this information. We are expl... HCI konda01actorcritic Actor-Critic Algorithms We propose and analyze a class of actor-critic algorithms for simulation-based optimization of a Markov decision process over a parameterized family of randomized stationary policies. These are two-time-scale algorithms in which the critic uses TD learning with a linear approximation architecture and the actor is updated in an approximate gradient direction based on information provided by the critic. We show that the features for the critic should span a subspace prescribed by the choice of parameterization of the actor. We conclude by discussing convergence properties and some open problems. 1 Introduction The vast majority of Reinforcement Learning (RL) [9] and Neuro-Dynamic Programming (NDP) [1] methods fall into one of the following two categories: (a) Actor-only methods work with a parameterized family of policies. The gradient of the performance, with respect to the actor parameters, is directly estimated by simulation, and the parameters are updated in a direction o... ML marchionini98interfaces Interfaces and Tools for the Library of Congress National Digital Library Program This paper describes a collaborative effort to explore user needs in a digital library, develop interface prototypes for a digital library, and suggest and prototype tools for digital librarians and users at the Library of Congress (LC). Interfaces were guided by an assessment of user needs and aimed to maximize interaction with primary resources and support both browsing and analytical search strategies. Tools to aid users and librarians in overviewing collections, previewing objects, and gathering results were created and serve as the beginnings of a digital librarian toolkit. The design process and results are described and suggestions for future work are offered. Digital Libraries (DL) offer new challenges to an emerging breed of digital librarians who must combine the principles and practices of information management with rapidly evolving technological developments to create new information products and services. This paper describes a collaborative effort to explore user needs ... HCI loyer99computing Computing and Comparing Semantics of Programs in Four-valued Logics The different semantics that can be assigned to a logic program correspond to different assumptions made concerning the atoms whose logical values cannot be inferred from the rules. Thus, the well founded semantics corresponds to the assumption that every such atom is false, while the Kripke-Kleene semantics corresponds to the assumption that every such atom is unknown. In this paper, we propose to unify and extend this assumption-based approach by introducing parameterized semantics for logic programs. The parameter holds the value that one assumes for all atoms whose logical values cannot be inferred from the rules. We work within Belnap's four-valued logic, and we consider the class of logic programs defined by Fitting. Following Fitting's approach, we define a simple operator that allows us to compute the parameterized semantics, and to compare and combine semantics obtained for different values of the parameter. The semantics proposed by Fitting corresponds to the value false. We also show that our approach captures and extends the usual semantics of conventional logic programs thereby unifying their computation. DB petrak95objectoriented An Object-Oriented Case-Based Learning System This thesis first gives an overview of the subfield of classification in the area of machine learning. The numerous variants of Case-Based Learning algorithms are compared according to what kind of data is processed, how knowledge and hypotheses are represented, and what kind of reasoning or learning is performed. The strengths and weaknesses of these learning methods are compared to each other and to other groups of learning methods. A modular object-oriented LISP environment, VIE-CBR2 is introduced, that implements a number of algorithms for Case-Based Learning. This system allows to easily combine preprogrammed learning algorithms. and provides a framework for simple integration of new learning algorithms and other components that make use of the basic system. ML borchers00pattern A Pattern Approach to Interaction Design To create successful interactive systems, user interface designers need to cooperate with developers and application domain experts in an interdisciplinary team. These groups, however, usually miss a common terminology to exchange ideas, opinions, and values. This paper presents an approach that uses pattern languages to capture this knowledge in software development, HCI, and the application domain. A formal, domain-independent definition of design patterns allows for computer support without sacrificing readability, and pattern use is integrated into the usability engineering lifecycle. As an example, experience from building an award-winning interactive music exhibit was turned into a pattern language, which was then used to inform follow-up projects and support HCI education. HCI saif01communication Communication Primitives for Ubiquitous Systems or RPC Considered Harmful RPC is widely used to access and modify remote state. Its procedural call semantics are argued as an efficient unifying paradigm for both local and remote access. Our experience with ubiquitous device control systems has shown otherwise. RPC semantics of a synchronous, blocking invocation on a statically typed interface are overly restrictive, inflexible, and fail to provide an efficient unifying abstraction for accessing and modifying state in ubiquitous systems. This position paper considers other alternatives and proposes the use of comvets (conditional, mobility aware events) as the unifying generic communication paradigm for such systems. Keywords: RPC, RMI, Events, Comvets, CORBA, Jini 1 Introduction Ubiquitous environments or active spaces are the next generation of device control networks. A user interacts with an active space by using novel interfaces like speech and gesture input [1] to control her environment, and the system interacts with the user using audio/video outpu... Agents 478335 Multi-User and Security Support for Multi-Agent Systems This paper discusses the requirements an agent system needs to be secure. In particular, the paper introduces a classification of modern distributed systems, and examines the delegation concept from a security point of view. After discussing the peculiar security and delegation issues present in distributed object systems, mobile agent systems and in multi agent systems, a case study is presented, describing the multi-user and security support that is being built into the JADE platform. Agents nambiar01benchmarking Benchmarking XML Management Systems: The XOO7 Way The effectiveness of existing XML query languages has been studied by many who focused on the comparison of linguistic features, implicitly reflecting the fact that most XML tools exist only on paper. In this paper, with a focus on efficiency and concreteness, we propose a pragmatic first step toward the systematic benchmarking of XML query processing platforms. We begin by identifying the necessary functionalities an XML data management system should support. We review existing approaches for managing XML data and the query processing capabilities of these approaches. We then compare three XML query benchmarks XMach-1, XMark and XOO7 and discuss the applicability, strengths and limitations of these benchmarks. We highlight the bias of these benchmarks towards the data centric view of XML and motivate our selection of XOO7 to extend with document centric queries. We complete XOO7 to capture the information retrieval capabilities of XML management systems. Finally we summarize our contributions and discuss future directions. DB 315017 dQUOB: Managing Large Data Flows Using Dynamic Embedded Queries The dQUOB system satisfies client need for specific information from high-volume data streams. The data streams we speak of are the flow of data existing during large-scale visualizations, video streaming to large numbers of distributed users, and high volume business transactions. We introduces the notion of conceptualizing a data stream as a set of relational database tables so that a scientist can request information with an SQL-like query. Transformation or computation that often needs to be performed on the data en-route can be conceptualized ascomputation performed onconsecutive views of the data, with computation associated with each view. The dQUOB system moves the query code into the data stream as a quoblet; as compiled code. The relational database data model has the significant advantage of presenting opportunities for efficient reoptimizations of queries and sets of queries. Using examples from global atmospheric modeling, we illustrate the usefulness of the dQUOB system. We carry the examples through the experiments to establish the viability of the approach for high performance computing with a baseline benchmark. We define a cost-metric of end-to-end latency that can be used to determine realistic cases where optimization should be applied. Finally, we show that end-to-end latency can be controlled through a probability assigned to a query that a query will evaluate to true. DB 301891 A Multi-version Approach to Conflict Resolution in Distributed Groupware Systems Groupware systems are a special class of distributed computing systems which support human-computer-human interaction. Real-time collaborative graphics editors allow a group of users to view and edit the same graphics document at the same time from geographically dispersed sites connected by communication networks. Resolving conflict access to shared objects is one of the core issues in the design of this type of systems. This paper proposes a novel distributed multi-version approach to conflict resolution. This approach aims to preserve the work concurrently produced by multiple users in the face of conflicts, and to minimize the number of object versions for accommodating combined effects of conflicting and compatible operations. Major technical contributions of this work include a formal specification of a unique combined effect for any group of conflicting and compatible operations, a distributed algorithm for incremental creation of multiple object versions, and a consistent objec... HCI dunlop00development Development and Evaluation of Clustering Techniques for Finding People Typically in a large organisation much expertise and knowledge is held informally within employees' own memories. When employees leave an organisation many documented links that go through that person are broken and no mechanism is usually available to overcome these broken links. This matchmaking problem is related to the problem of finding potential work partners in a large and distributed organisation. This paper reports a comparative investigation into using standard information retrieval techniques to group employees together based on their web pages. This information can, hopefully, be subsequently used to redirect broken links to people who worked closely with a departed employee or used to highlight people, say in different departments, who work on similar topics. The paper reports the design and positive results of an experiment conducted at Ris National Laboratory comparing four different IR searching and clustering approaches using real users' we... IR 292710 Use Case Maps as a Feature Description Notation . We propose Use Case Maps (UCMs) as a notation for describing features. UCMs capture functional requirements in terms of causal scenarios bound to underlying abstract components. This particular view proved very useful in the description of a wide range of reactive and telecommunications systems. This paper presents some of the most interesting constructs and benefits of the notation in relation to a question on a User Requirements Notation recently approved by ITU-T Study Group 10, which will lead to a new Recommendation by 2003. Tool support, current research on UCMs, and related notations are also discussed. 1 Introduction The modeling of reactive systems requires an early emphasis on behavioral aspects such as interactions between the system and the external world (including the users), on the cause-to-e#ect relationships among these interactions, and on intermediate activities performed by the system. Scenarios are particularly good at representing such aspects so that various ... DB klein99computer Computer Response to User Frustration Use of computer technology often has unpleasant side effects, some of which are strong, negative emotional states that arise in humans during interaction with computers. Frustration, confusion, anger, anxiety and similar emotional states can affect not only the interaction itself, but also productivity, learning, social relationships, and overall well-being. This thesis presents the idea of designing human-computer interaction systems to actively support human users in their ability to regulate, manage, and recover from their own negative emotional states, particularly frustration. This document describes traditional theoretical strategies for emotion regulation, the design of a human-computer interaction agent built by the author to actively help relieve frustration, and an evaluation that shows the effectiveness of the agent. A study designed to test this agent was conducted: A system was built that elicits frustration in human subjects. The interaction agent then initiated several social, emotional-content feedback strategies with some of the subjects, in an effort to help relieve their emotional state. These strategies were designed to provide many of the same cues that skilled, human listeners employ when helping relieve strong, negative emotions in others. Two control groups were exposed to the same frustrating stimuli, one of which was given no emotional support at all; the other enabled subjects to report problems and "vent" at the computer. Subsequent behavior was then observed, and self-report data was collected. Behavioral results showed the agent was significantly more effective than the two controls in helping relieve frustration levels in subjects. These results demonstrate that strategic, social, emotional-content interaction with a computer by users who ... HCI faensen01hermes Hermes - A Notification Service for Digital Libraries The high publication rate of scholarly material makes searching and browsing an inconvenient way to keep oneself up-todate. Instead of being the active part in information access, researchers want to be notified whenever a new paper in one's research area is published. IR slaughter00open Open video: A framework for a test collection This paper provides a framework for such a test collection and describes the Open Video Project that has begun to develop a test collection based on this framework. The proposed test collection is meant to be used to study a wide range of problems, such as tests of algorithms for creating surrogates for video content or interfaces that display result sets from queries. An important challenge in developing such a collection is storing and distributing video objects. This paper is meant to layout video management issues that may influence distributed storage solutions. More specifically, this paper describes the first phase for creating the test collection, sets guidelines for building the collection, and serves as a basis for discussion to inform subsequent phases and invite research community involvement. 2000 Academic Press 1. Introduction It is inevitable that the technical limitations that impede widespread usage of video libraries will dimi IR malec00implementing Implementing Teams of Agents Playing Simulated Robot Soccer . This article is intended to present an overview of the issues related to implementing teams of cooperating agents playing simulated robot soccer. First we introduce the concept of robot soccer and the simulated environment. Then we discuss why the (simulated) robot soccer is an interesting application area from the point of view of robotics and articial intelligence. The main part of the paper contains a discussion of agent architectures, both from theoretical and practical point of view. Later we discuss how to combine individual agents into teams having common strategies and goals. We also discuss learning, both on individual and team levels. 1 Introduction Robot soccer is a growing area of interest for the robotics and articial intelligence communities. There are many reasons for that. The main one is the complexity of the domain together with a set of well-dened rules governing the behaviour of the agents in this domain. The domain is suited for experiments both in ... Agents 384644 Formal Concepts of Learning Systems Validation in Use In the problem area of evaluating complex software systems, there are two distinguished areas of research, development, and application identified by the two buzzwords validation and verification, respectively. From the perspective adopted by the authors (cf. (O'Keefe & O'Leary 1993), e.g.), verification is usually more formally based and, thus, can be supported by formal reasoning tools like theorem provers, for instance. The scope of verification approaches is limited by the difficulty of finding a sufficiently complete formalization to built upon. In paramount realistic problem domains, validation seems to be more appropriate, although it is less stringent in character and, therefore, validation results are often less definite. The aim of this paper is to exemplify a validation approach based on a clear and thoroughly formal theory. In this way, validation and verification should be brought closer to each other, for the benefit of a concerted action towards depend... ML punin01web Web Usage Mining - Languages and Algorithms We propose two new XML applications, XGMML and LOGML. XGMML is a graph description language and LOGML is a web-log report description language. We generate a web graph in XGMML format for a web site using the web robot of the WWWPal system (developed for web visualization and organization). We generate web-log reports in LOGML format for a web site from web log files and the web graph. In this paper, we further illustrate the usefulness of these two XML applications with a web data mining example. Moreover, we show the simplicity with which this mining algorithm can be specified and implemented efficiently using our two XML applications. We provide sample results, namely frequent patterns of users in a web site, with our web data mining algorithm. IR li01minimizing Minimizing View Sets without Losing Query-Answering Power The problem of answering queries using views has been studied extensively, due to its relevance in a wide variety of data-management applications. In these applications, we often need to select a subset of views to maintain, due to limited resources. In this paper, we show that traditional query containment is not a good basis for deciding whether or not a view should be selected. Instead, we should minimize the view set without losing query-answering power. To formalize this notion, we rst introduce the concept of "p-containment." That is, a view set V is p-contained in another view set W, if W can answer all the queries that can be answered by V. We show that p-containment and the traditional query containment are not related; i.e., one does not imply the other. We then discuss how to minimize a view set while retaining its query-answering power. We develop the idea further by considering p-containment of two view sets with respect to a given set of queries, and consider their relationship in terms of maximally-contained rewritings of queries using the views. DB killijian01towards Towards Group Communication for Mobile Participants (Extended Abstract) Group communication will undoubtedly be a useful paradigm for many applications of wireless networking in which reliability and timeliness are requirements. Moreover, location awareness is clearly central to mobile applications such as traffic management and smart spaces. In this paper, we introduce our definition of proximity groups in which group membership depends on location and then discuss some requirements for a group membership management service suitable for proximity groups. We describe a novel approach to efficient coverage estimation, giving applications feedback on the proportion of the area of interest covered by a proximity group, and also discuss our approach to partition anticipation. HCI mitzenmacher01estimating Estimating Resemblance of MIDI Documents Abstract. Search engines often employ techniques for determining syntactic similarity of Web pages. Such a tool allows them to avoid returning multiple copies of essentially the same page when a user makes a query. Here we describe our experience extending these techniques to MIDI music files. The music domain requires modification to cope with problems introduced in the musical setting, such as polyphony. Our experience suggests that when used properly these techniques prove useful for determining duplicates and clustering databases in the musical setting as well. 1 IR piater00distinctive Distinctive Features Should Be Learned Most existing machine vision systems perform recognition based on a xed set of hand-crafted features, geometric models, or eigen-subspace decomposition. Drawing from psychology, neuroscience and intuition, we show that certain aspects of human performance in visual discrimination cannot be explained by any of these techniques. We argue that many practical recognition tasks for articial vision systems operating under uncontrolled conditions critically depend on incremental learning. Loosely motivated by visuocortical processing, we present feature representations and learning methods that perform biologically plausible functions. The paper concludes with experimental results generated by our method. 1 Introduction How exible are the representations for visual recognition, encoded by the neurons of the human visual cortex? Are they predetermined by a xed developmental schedule, or does their development depend on their stimulation? Does their development cease at some poin... ML 232670 Interaction between Path and Type Constraints XML [7], which is emerging as an important standard for data exchange on the World-Wide Web, highlights the importance of semistructured data. Although the XML standard itself does not require any schema or type system, a number of proposals [6, 17, 19] have been developed that roughly correspond to data definition languages. These allow one to constrain the structure of XML data by imposing a schema on it. These and other proposals also advocate the need for integrity constraints, another form of constraints that should, for example, be capable of expressing inclusion constraints and inverse relationships. The latter have recently been studied as path constraints in the context of semistructured data [4, 9]. It is likely that future XML proposals will involve both forms of constraints, and it is therefore appropriate to understand the interaction between them. This paper investigates that interaction. In particular it studies constraint implication problems, which are important both i... DB jennings01aspects Aspects of Network Edge Intelligence Is it the case that the migration of intelligence from the core of networks to the periphery is simply a function of the IP protocols? Or are there more fundamental forces at work? This report addresses this issue from the perspective of core network protocols, mobile networking and the emerging embedded Internet. It identifies the forces at work, and concludes that the long term trends are driven by more than simply the end-to-end argument. If the logical progression of this trend is that intelligence will migrate to the embedded Internet, then a new type of peripheral intelligence may form the basis for further progress. The report concludes by identifying the challenges for the embedded Internet in constructed environments. Keywords: IP networking, intelligence, architecture, smart spaces Aspects of Network Edge Intelligence I. HCI 92649 Robustness of Case-Initialized Genetic Algorithms We investigate the robustness of Case Initialized Genetic AlgoRithm (CIGAR) systems with respect to problem indexing. When confronted with a series of similar problems CIGAR stores potential solutions in a case-base or an associative memory and retrieves and uses these solutions to help improve a genetic algorithm 's performance over time. Defining similarity among the problems, or indexing, is key to performance improvement. We study four indexing schemes on a class of simple problems and provide empirical evidence of CIGAR's robustness to imperfect indexing. ML ciravegna00learning Learning to Tag for Information Extraction from Text . LearningPINOCCHIO is an algorithm for adaptive information extraction. It learns template filling rules that insert SGML tags into texts. LearningPINOCCHIO is based on a covering algorithm that learns rules by bottom-up generalization of instances in a tagged corpus. It has been tested on three scenarios in informal domains in two languages (Italian and English). Experiments report excellent results with respect to the current state of the art. 1 Introduction By general agreement the main barriers to wide use and commercialization of IE are the difficulties in adapting systems to new applications and domains. In the last years there has been increasing interest in applying machine learning to Information Extraction from text [13, 3, 9, 1, 11]. In particular there is an increasing interest in the application of adaptive IE to Web pages [12, 10] and to informal domains (rental ads, email messages, etc.) [15, 8, 2] for building fully automated systems. This is due from one side to the... ML 457459 A Natural Interface to a Virtual Environment through Computer Vision-estimated Pointing Gestures . This paper describes the development of a natural interface to a virtual environment. The interface is through a natural pointing gesture and replaces pointing devices which are normally used to interact with virtual environments. The pointing gesture is estimated in 3D using kinematic knowledge of the arm during pointing and monocular computer vision. The latter is used to extract the 2D position of the user's hand and map it into 3D. Off-line tests of the system show promising results with an average errors of 76mm when pointing at a screen 2m away. The implementation of a real time system is currently in progress and is expected to run with 25Hz. 1 HCI hannebauer99rapid Rapid Concurrent Software Engineering in Competitive Situations This article is an experience report on the evolutionary development process of AT Humboldt, a multi agent system which has become World Champion 1997 and Vice World Champion 1998 of RoboCup simulator league. It details why the artifical soccer initiative RoboCup is a tempting domain for rapid concurrent software engineering. Both the development processes in 1997 and 1998 are described, compared and evaluated. Lessons learned for development projects in distributed control conclude this report. 1 Introduction In this article the project managers describe the evolutionary development process of the software project AT (AgentTeam) Humboldt, which has become World Champion 1997 and Vice World Champion 1998 in the simulator league of the artifical soccer contest RoboCup ([10]). The RoboCup initiative recently gets more and more popular among scientists in robotics, distributed systems and distributed artificial intelligence because of its strong competitive character and tight resource b... Agents bredin99economic Economic Markets as a Means of Open Mobile-Agent Systems Mobile-agent systems have gained popularity in use because they ease the application design process by giving software engineers greater flexibility. Although the value of any network is dependent on both the number of users and the number of sites participating in the network, there is little motivation for systems to donate resources to arbitrary agents. We propose to remedy the problem by imposing an economic market on mobile-agent systems where agents purchase resources from host sites and sell services to users and other agents. Host sites accumulate revenues, which are distributed to users to be used to launch more agents. We argue for the use of markets to regulate mobile-agent systems and discuss open issues in implementing market-based mobile-agent systems. 1 Introduction One of the more recent items in a network programmer's tool box is code mobility. The technique is becoming more common in applications programming, network management [BPW98], video conferencing [BPR98], so... Agents oliver98graphical Graphical Models for Recognizing Human Interactions We describe a real-time computer vision and machine learning system for modeling and recognizing human actions and interactions. Two different domains are explored: recognition of two-handed motions in the martial art 'Tai Chi', and multiple-person interactions in a visual surveillance task. Our system combines top-down with bottom-up information using a feedback loop, and is formulated with a Bayesian framework. Two different graphical models (HMMs and Coupled HMMs) are used for modeling both individual actions and multiple-agent interactions, and CHMMs are shown to work more efficiently and accurately for a given amount of training. Finally, to overcome the limited amounts of training data, we demonstrate that `synthetic agents' (Alife-style agents) can be used to develop flexible prior models of the person-to-person interactions. 1 INTRODUCTION We describe a real-time computer vision and machine learning system for modeling and recognizing human behaviors in two different scenari... Agents 509627 Adaptive Load Sharing for Network Processors A novel scheme for processing packets in a router is presented, which provides for load sharing among multiple network processors distributed within the router. It is complemented by a feedback control mechanism designed to prevent processor overload. Incoming traffic is scheduled to multiple processors based on a deterministic mapping. The mapping formula is derived from the robust hash routing (also known as the highest random weight - HRW) scheme, introduced in K.W. Ross, IEEE Network, 11(6), 1997, and D.G. Thaler et al., IEEE Trans. Networking, 6(1), 1998. No state information on individual flow mapping needs to be stored, but for each packet, a mapping function is computed over an identifier vector, a predefined set of fields in the packet. An adaptive extension to the HRW scheme is provided in order to cope with biased traffic patterns. We prove that our adaptation possesses the minimal disruption property with respect to the mapping and exploit that property in order to minimize the probability of flow reordering. Simulation results indicate that the scheme achieves significant improvements in processor utilization. A higher number of router interfaces can thus be supported with the same amount of processing power. I. IR chan98approximate Approximate Nearest Neighbor Queries Revisited This paper proposes new methods to answer approximate nearest neighbor queries on a set of n points in d-dimensional Euclidean space. For any fixed constant d, a data structure with O(" (1\Gammad)=2 n log n) preprocessing time and O(" (1\Gammad)=2 log n) query time achieves approximation factor 1 + " for any given 0 ! " ! 1; a variant reduces the "-dependence by a factor of " \Gamma1=2 . For any arbitrary d, a data structure with O(d 2 n log n) preprocessing time and O(d 2 log n) query time achieves approximation factor O(d 3=2 ). Applications to various proximity problems are discussed. 1 Introduction Let P be a set of n point sites in d-dimensional space IR d . In the well-known post office problem, we want to preprocess P into a data structure so that a site closest to a given query point q (called the nearest neighbor of q) can be found efficiently. Distances are measured under the Euclidean metric. The post office problem has many applications within computational... ML 138451 A Geometric Framework for Specifying Spatiotemporal Objects We present a framework for specifying spatiotemporal objects using spatial and temporal objects, and a geometric transformation. We define a number of classes of spatiotemporal objects and study their closure properties. 1 Introduction Many natural or man-made phenomena have both a spatial and a temporal extent. Consider for example, a forest fire or property histories in a city. To store information about such phenomena in a database one needs appropriate data modeling constructs. We claim that a new concept, spatiotemporal object, is necessary. In this paper, we introduce a very general framework for specifying spatiotemporal objects. To define a spatiotemporal object we need a spatial object, a temporal object, and a continuous geometric transformation (specified using a parametric representation) that determines the image of the spatial object at different time instants belonging to the temporal object. In this framework, a number of classes of spatiotemporal objects arise quite ... DB benetti01sidesigner SI-Designer: a tool for intelligent integration of information SI-Designer (Source Integrator Designer) is a designer support tool for semi � automatic integration of heterogeneous sources schemata (relational, object and semi � structured sources); it has been implemented within the MOMIS project and it carries out integration following a semantic approach which uses intelligent Description Logics-based techniques, clustering techniques and an extended ODMG-ODL language, �������� � , to represent schemata, extracted, integrated information. Starting from the sources ’ �������� � descriptions (local schemata) SI-Designer supports the designer in the creation of an integrated view of all the sources (global schema) which is expressed in the same �������� � language. We propose SI-Designer as a tool to build virtual catalogs in the E-Commerce environment. 1. DB 324100 Information-Theoretic Learning This chapter seeks to extend the ubiquitous mean-square error criterion (MSE) to cost functions that include more information about the training data. Since the learning process ultimately should transfer the information carried in the data samples onto the system's parameters, a natural goal is to find cost functions that directly manipulate information. Hence the name informationtheoretic learning (ITL). In order to be useful, ITL should be independent of the learning machine architecture, and require solely the availability of the data, i.e. it should not require a priori assumptions about the data distributions. The chapter presents our current efforts to develop ITL criteria based on the integration of nonparametric density estimators with Renyi's quadratic entropy definition. As a motivation we start with an application of the MSE to manipulate information using the nonlinear characteristics of the learning machine. This section illustrates the issues faced when we attempt to use... IR wu99firstorder First-Order Polynomial Based Theorem Proving Introduction The Boolean ring or first-order polynomial based theorem proving began with the work of Hsiang (1982, 1985). Hsiang extended the idea of using Boolean polynomials to represent propositional formulae to the case of first-order predicate calculus. Based on the completion procedure of Knuth and Bendix (1970), the N-strategy was proposed. Later on, by imitating the framework of Buchberger 's algorithm to compute the Grobner bases of polynomial ideals (Buchberger 1985), Kapur and Narendran (1985) developed another approach which is also referred to as the Grobner basis method. One obvious advantage of using Boolean polynomials is that every propositional formula has a unique representation, and sometimes it is easy to be generalized to some non-classical logic systems (Chazarain et al. 1991; Wu and Tan 1994). Stimulated by them, some approaches and results have been reported (Bachmair and Dershowitz 1987; Dietrich 1986; Kapur and Zhang 1989; Wu and Liu 1998; Zhang 198 DB turney00learning Learning Algorithms for Keyphrase Extraction Many academic journals ask their authors to provide a list of about five to fifteen keywords, to appear on the first page of each article. Since these key words are often phrases of two or more words, we prefer to call them keyphrases. There is a wide variety of tasks for which keyphrases are useful, as we discuss in this paper. We approach the problem of automatically extracting keyphrases from text as a supervised learning task. We treat a document as a set of phrases, which the learning algorithm must learn to classify as positive or negative examples of keyphrases. Our first set of experiments applies the C4.5 decision tree induction algorithm to this learning task. We evaluate the performance of nine different configurations of C4.5. The second set of experiments applies the GenEx algorithm to the task. We developed the GenEx algorithm specifically for automatically extracting keyphrases from text. The experimental results support the claim that a custom-designed algorithm (GenEx)... IR sawhney99nomadic Nomadic Radio: Scaleable and Contextual Notification for Wearable Audio Messaging Mobile workers need seamless access to communication and information services on portable devices. However current solutions overwhelm users with intrusive and ambiguous notifications. In this paper, we describe scaleable auditory techniques and a contextual notification model for providing timely information, while minimizing interruptions. User's actions influence local adaptation in the model. These techniques are demonstrated in Nomadic Radio, an audio-only wearable computing platform. HCI 369356 Object-Based Multimedia Content Description Schemes and Applications for MPEG-7 In this paper, we describe description schemes (DSs) for image, video, multimedia, home media, and archive content proposed to the MPEG-7 standard. MPEG-7 aims to create a multimedia content description standard in order to facilitate various multimedia searching and filtering applications. During the design process, special care was taken to provide simple but powerful structures that represent generic multimedia data. We use the eXtensible Markup Language (XML) to illustrate and exemplify the proposed DSs because of its interoperability and flexibility advantages. The main components of the image, video, and multimedia description schemes are object, feature classification, object hierarchy, entity-relation graph, code downloading, multi-abstraction levels, and modality transcoding. The home media description instantiates the former DSs proposing the 6-W semantic features for objects, and 1-P physical and 6-W semantic object hierarchies. The archive description scheme aims to describ... IR 443403 Error-Tolerant Agents . The use of agents in today's Internet world is expanding rapidly. Yet, agent developers Agents 286829 Approximate Query Translation across Heterogeneous Information Sources (Extended Version) In this paper we present a mechanism for approximately translating Boolean query constraints across heterogeneous information sources. Achieving the best translation is challenging because sources support different constraints for formulating queries, and often these constraints cannot be precisely translated. For instance, a query [score ? 8] might be "perfectly" translated as [rating ? 0.8] at some site, but can only be approximated as [grade = A] at another. Unlike other work, our general framework adopts a customizable "closeness" metric for the translation that combines both precision and recall. Our results show that for query translation we need to handle interdependencies among both query conjuncts as well as disjuncts. As the basis, we identify the essential requirements of a rule system for users to encode the mappings for atomic semantic units. Our algorithm then translates complex queries by rewriting them in terms of the semantic units. We show that, un... DB roobaert00directsvm DirectSVM: A Fast And Simple Support Vector Machine Perceptron . We propose a simple implementation of the Support Vector Machine (SVM) for pattern recognition, that is not based on solving a complex quadratic optimization problem. Instead we propose a simple, iterative algorithm that is based on a few simple heuristics. The proposed algorithm nds high-quality solutions in a fast and intuitively-simple way. In experiments on the COIL database, on the extended COIL database and on the Sonar database of the UCI Irvine repository, DirectSVM is able to nd solutions that are similar to these found by the original SVM. However DirectSVM is able to nd these solutions substantially faster, while requiring less computational resources than the original SVM. INTRODUCTION Support Vector Machines (SVMs) belong to the best-performing learning algorithms available. They have produced remarkable performance in a number of dicult learning tasks without requiring prior knowledge. We mention amongst others the following examples in pattern recognition: handwr... ML 119642 Bias, Variance, and Error Correcting Output Codes for Local Learners : This paper focuses on a bias variance decomposition analysis of a local learning algorithm, the nearest neighbor classifier, that has been extended with error correcting output codes. This extended algorithm often considerably reduces the 0-1 (i.e., classification) error in comparison with nearest neighbor (Ricci & Aha, 1997). The analysis presented here reveals that this performance improvement is obtained by drastically reducing bias at the cost of increasing variance. We also show that, even in classification problems with few classes (m5), extending the codeword length beyond the limit that assures column separation yields an error reduction. This error reduction is not only in the variance, which is due to the voting mechanism used for error-correcting output codes, but also in the bias. Keywords: Case-based learning, classification, error-correcting output codes, bias and variance Email: ricci@irst.itc.it, aha@aic.nrl.navy.mil Phone: ++39 461 314334 FAX: ++39 461 302040 Bi... ML 457569 Schema Evolution in Heterogeneous Database Architectures, A Schema Transformation Approach In previous work we have a developed general framework to support schema transformation and integration in heterogeneous database architectures. The framework consists of a hypergraph-based common data model and a set of primitive schema transformations defined for this model. Higher-level common data models and primitive schema transformations for them can be defined in terms of this lower-level model. A key feature of the framework is that both primitive and composite schema transformations are automatically reversible. We have shown in earlier work how this allows automatic query translation from a global schema to a set of source schemas. In this paper we show how our framework also readily supports evolution of source schemas, allowing the global schema and the query translation pathways to be easily repaired, as opposed to having to be regenerated, after changes to source schemas. 1 DB neaves98dynamic Dynamic Connection of Wearable Computers to Companion Devices Using Near-Field Radio Hewlett-Packard Laboratories, Bristol and the University of Bristol Department of Computer Science are engaged in an initiative to explore the design, technology and use of wearable computers. We describe a way of connecting a wearable computer to companion devices such as displays or cameras using near-field radio technology. The shortrange nature of near-field radio allows relatively high data rates (300 kbps -- 1Mbit), low power consumption and the interpretation of gestures as configuration requests. Keywords: Near-field radio, dynamic connectivity. INTRODUCTION We are particularly interested in communication technologies that exhibit low-power, short range (up to 1 foot) and modest data rates (300 kbps -- 1 Mbs). The action of picking up a companion device (such as a display) establishes the communication link due to the very short range. An important aspect of a suitable communication technology is that the user is not required to touch an electrode and therefore handling of ... HCI 147739 SIC: Satisfiability Checking for Integrity Constraints SIC is an interactive prototype to assist in the design of finitely satisfiable integrity constraints. Thus SIC addresses the constraint satisfiability problem during the schema design phase of a database. SIC combines two systems, a reasoning component and an interactive visual interface. This paper outlines the functionality of both components and the theoretical background and implementation aspects of the reasoning component. DB stocky02conveying Conveying Routes: Multimodal Generation and Spatial Intelligence In Embodied Conversational Agents In creating an embodied conversational agent (ECA) capable of conveying routes, it is necessary to understand how to present spatial information in an effective and natural manner. When conveying routes to someone, a person uses multiple modalities – e.g., speech, gestures, and reference to a map – to present information, and it is important to know precisely how these modalities are coordinated. With an understanding of how humans present spatial intelligence to give directions, it is then possible to create an ECA with similar capabilities. Two empirical studies were carried out to observe natural human-to-human direction-giving interactions. From the results, a direction-giving model was created, and then implemented in the MACK (Media Lab Autonomous Conversational Kiosk) system. HCI paradiso99cybershoe The CyberShoe: A Wireless Multisensor Interface for a Dancer's Feet : As a bridge between our interest in Wearable Computer systems and new performance interfaces for digital music, we have built a highly instrumented pair of sneakers for interactive dance. These shoes each measure 16 different, continuous parameters expressed by each foot and are able to transmit them wirelessly to a base station placed well over 30 meters away, updating all values up to 60 times per second. This paper describes our system, illustrates its performance, and outlines a few musical mappings that we have created for demonstrations in computer-augmented dance. ____________________________________ Electronic sensors have been incorporated into footwear for several different applications over the last several years. Employing force-sensing resistor arrays or pixelated capacitive sensing, insoles with very dense pressure sampling have been developed for research at the laboratories of footwear manufacturers and pediatric treatment facilities (Cavanaugh, et. al., 1992). Alth... HCI nodine98overview An Overview of Active Information Gathering in InfoSleuth InfoSleuth is a system of collaborating software agents that can be configured to perform many different information management activities in a distributed environment. InfoSleuth agents advertise semantic constraints about themselves to InfoSleuth brokers using a global domain ontology. When queried, a broker reasons over these constraints to determine the minimal set of agents that can provide a solution to the query. InfoSleuth's architecture is based on a generic agent shell that provides basic agent communication behaviors over a subset of Knowledge Query Manipulation Language. Individual agents are subclasses of this generic shell that provide specific kinds of functionality. InfoSleuth agents perform a number of complex query activities that require resolving ontology-based queries over dynamically changing, distributed, heterogeneous resources, including distributed query, location-independent single-resource updates, event monitoring by means of subscription/notification servi... DB 110962 Logical Transactions and Serializability The concept of logic databases can serve as a clear and expressive foundation of various information systems. However, the classical logic language only refers to a single database state, although in modern information systems it is necessary to take the dynamics into account. For this purpose, several update languages were proposed, among them [1, 3, 4, 8, 9, 11], which allow to specify complex transitions from one database state to another. From the evaluation point of view, a complex state transition can and should be considered as a transaction. Up to now, the isolation property of transactions has been poorly addressed in the context of logic update languages, although it is an important problem even for classical, sequential transactions (see [2] for instance). In this paper we investigate how the serializability of logical transactions can be supported and what this means for the implementation of a transaction manager. 1 Introduction and Motivation In [11, 12] we propose an up... DB agarwal01time Time Responsive Indexing Schemes for Moving Points We develop new indexing schemes for storing a set of points in one or two dimensions, each moving along a linear trajectory, so that a range query at a given future time t q can be answered efficiently. The novel feature of our indexing schemes is that the number of I/Os required to answer a query depends not only on the size of the data set and on the number of points in the answer but also on the difference between t q and the current time; queries close to the current time are answered fast, while queries that are far away in the future or in the past may take more time. Center for Geometric Computing, Department of Computer Science, Duke University, Durham, NC 27708, USA. Supported in part by Army Research Office MURI grant DAAH04-96-1-0013, by a Sloan fellowship, by NSF grants EIA--9870724, EIA--997287, and CCR--9732787, and by grant from the U.S.-Israeli Binational Science Foundation. Email: pankaj@cs.duke.edu. y Center for Geometric Computing, Department of Computer Sci... DB 31326 Inference and Learning in Hybrid Bayesian Networks We survey the literature on methods for inference and learning in Bayesian Networks composed of discrete and continuous nodes, in which the continuous nodes have a multivariate Gaussian distribution, whose mean and variance depends on the values of the discrete nodes. We also briefly consider hybrid Dynamic Bayesian Networks, an extension of switching Kalman filters. This report is meant to summarize what is known at a sufficient level of detail to enable someone to implement the algorithms, but without dwelling on formalities. 1 1 Introduction We discuss Bayesian networks (BNs [Jen96]) in which each node is either discrete or continuous, scalar or vector-valued, and in which the joint distribution over all the nodes is Conditional Gaussian (CG) [LW89, Lau92] i.e., for each instantiation i of the discrete nodes Y, the distribution over the continuous nodes X has the form f(xjY = i) = N (x; ~¯(i); \Sigma(i)), where N () represents a multivariate Gaussian (MVG) or Normal density. (Note... ML 539227 EQUIP: a Software Platform for Distributed Interactive Systems EQUIP is a new software platform designed and engineered to support the development and deployment of distributed interactive systems, such as mixed reality user interfaces that combine distributed input and output devices to create a coordinated experience. EQUIP emphasises: cross-language development (currently C++ and Java), modularisation, extensibility, interactive performance, and heterogeneity of devices (from handheld devices to large servers and visualisation machines) and networks (including both wired and wireless technologies). A key element of EQUIP is its shared data service, which combines ideas from tuplespaces, general event systems and collaborative virtual environments. This data service provides a uniquely balanced treatment of state and event-based communication. It also supports distributed computation -- through remote class loading -- as well as passive data distribution. EQUIP has already been used in several projects within the EQUATOR Interdisciplinary Research Collaboration (IRC) in the UK, and is freely available in source form (currently known to work on Windows, IRIX and MacOS-X platforms). HCI 29319 Optimising Propositional Modal Satisfiability for Description Logic Subsumption . Effective optimisation techniques can make a dramatic difference in the performance of knowledge representation systems based on expressive description logics. Because of the correspondence between description logics and propositional modal logic many of these techniques carry over into propositional modal logic satisfiability checking. Currently-implemented representation systems that employ these techniques, such as FaCT and DLP, make effective satisfiable checkers for various propositional modal logics. 1 Introduction Description logics are a logical formalism for the representation of knowledge about individuals and descriptions of individuals. Description logics represent and reason with descriptions similar to "all people whose friends are both doctors and lawyers" or "all people whose children are doctors or lawyers or who have a child who has a spouse". The computations performed by systems that implement description logics are based around determining whether one descriptio... AI triesch98gesture A Gesture Interface for Human-Robot-Interaction We present a person-independent gesture interface implemented on a real robot which allows the user to give simple commands ,e.g., how to grasp an object and where to put it. The gesture analysis relies on realtime tracking of the user's hand and a re\Thetaned analysis of the hand's shape in the presence of varying complex backgrounds. 1. Introduction Robots of the future will interact with humans in a natural way. They will understand spoken and gestural commands and will articulate themselves by speech and gesture. We are especially interested in gestural interfaces for robots operating in uncontrolled real world environments. This imposes several constraints on human-robot-interaction as a special case of human-computer-interaction: 1. The robot visual system must cope with variable and possibly complex backgrounds. A system requiring uniform background is not exible enough for real world applications. 2. The system must be person independent. Many users should be able to operate ... AI galhardas01improving Improving Data Cleaning Quality using a Data Lineage Facility The problem of data cleaning, which consists of removing inconsistencies and errors from original data sets, is well known in the area of decision support systems and data warehouses. However, for some applications, existing ETL (Extraction Transformation Loading) and data cleaning tools for writing data cleaning programs are insufficient. One important challenge with them is the design of a data flow graph that effectively generates clean data. A generalized difficulty is the lack of explanation of cleaning results and user interaction facilities to tune a data cleaning program. This paper presents a solution to handle this problem by enabling users to express user interactions declaratively and tune data cleaning programs. 1 DB 330661 Maude: Specification and Programming in Rewriting Logic Maude is a high-level language and a high-performance system supporting executable specification and declarative programming in rewriting logic. Since rewriting logic contains equational logic, Maude also supports equational specification and programming in its sublanguage of functional modules and theories. The underlying equational logic chosen for Maude is membership equational logic, that has sorts, subsorts, operator overloading, and partiality definable by membership and equality conditions. Rewriting logic is reflective, in the sense of being able to express its own metalevel at the object level. Reflection is systematically exploited in Maude endowing the language with powerful metaprogramming capabilities, including both user-definable module operations and declarative strategies to guide the deduction process. This paper explains and illustrates with examples the main concepts of Maude's language design, including its underlying logic, functional, system and object-oriented modules, as well as parameterized modules, theories, and views. We also explain how Maude supports reflection, metaprogramming and internal strategies. The paper outlines the principles underlying the Maude system implementation, including its semicompilation techniques. We conclude with some remarks about applications, work on a formal environment for Maude, and a mobile language extension of Maude. DB poupyrev98virtual Virtual Notepad: Handwriting in Immersive VR We present Virtual Notepad, a collection of interface tools that allows the user to take notes, annotate documents and input text using a pen, while still immersed in virtual environments (VEs). Using a spatially-tracked, pressure-sensitive graphics tablet, pen and handwriting recognition software, Virtual Notepad explores handwriting as a new modality for interaction in immersive VEs. This paper reports details of the Virtual Notepad interface and interaction techniques, discusses implementation and design issues, reports the results of initial evaluation and overviews possible applications of virtual handwriting. 1. Introduction Writing is a ubiquitous everyday activity. We jot down ideas and memos, scribble comments in the margins of a book or an article, annotate blueprints and design plans. Using computers, we type documents, complete forms and enter database queries. However, writing, taking notes or entering text in immersive VEs is almost impossible. Cut off from conventional... DB dzeroski98detecting Detecting traffic problems with ILP Expert systems for decision support have recently been successfully introduced in road transport management. These systems include knowledge on traffic problem detection and alleviation. The paper describes experiments in automated acquisition of knowledge on traffic problem detection. The task is to detect road sections where a problem has occured (critical sections) from sensor data. It is necessary to use inductive logic programming (ILP) for this purpose as relational background knowledge on the road network is essential. In this paper, we apply three state-of-the art ILP systems to learn how to detect traffic problems and compare their performance to the performance of a propositional learning system on the same problem. 1 Introduction Expert systems for decision support have recently been successfully introduced in road transport management. Some of the proposals in this direction are TRYS [5], KITS [4] and ARTIST [9]. From a general perspective, the goal of a real time traffi... DB uhrmacher00modeling Modeling And Simulation Of Mobile Agents Agent-oriented software implies the realization of software components, which are mobile, autonomous, and solve problems by creating new software components during run-time, moving between locations, initiating or joining groups of other software components. Modeling and simulating those multiagent systems requires specific mechanisms for variable structure modeling. JAMES, a Java-Based Agent Modeling Environment for Simulation, realizes variable structure models including mobility from the perspective of single autonomous agents. JAMES itself is based on parallel DEVS and adopts its abstract simulator model. Simulation takes place as a sending of messages between concurrently active and locally distributed entities which reflect the model's current structure. Thus, modeling and simulation are coined equally by an agent-based perspective. 1 Introduction The definition of agents subsumes a multitude of different facets [30]. Agents are reactive, deliberative or combine reactive with ... Agents 513722 Parameterized Logic Programs where Computing Meets Learning Abstract. In this paper, we describe recent attempts to incorporate learning into logic programs as a step toward adaptive software that can learn from an environment. Although there are a variety of types of learning, we focus on parameter learning of logic programs, one for statistical learning by the EM algorithm and the other for reinforcement learning by learning automatons. Both attempts are not full- edged yet, but in the former case, thanks to the general framework and an e cient EM learning algorithm combined with a tabulated search, we have obtained very promising results that open up the prospect of modeling complex symbolic-statistical phenomena. 1 ML 25809 Object-relational Queries into Multidimensional Databases with the Active Data Repository As computational power and storage capacity increase, processing and analyzing large volumes of multi-dimensional datasets play an increasingly important role in many domains of scientific research. Scientific applications that make use of very large scientific datasets have several important characteristics: datasets consist of complex data and are usually multi-dimensional; applications usually retrieve a subset of all the data available in the dataset; various application-specific operations are performed on the data items retrieved. Such applications can be supported by object-relational database management systems (OR-DBMSs). In addition to providing functionality to define new complex datatypes and user-defined functions, an OR-DBMS for scientific datasets should contain runtime support that will provide optimized storage for very large datasets and an execution environment for user-defined functions involving expensive operations. In this paper we describe an infrastructure, t... DB hustadt00verification Verification within the KARO Agent Theory Abstract. This paper discusses automated reasoning in the KARO framework. The KARO framework accommodates a range of expressive modal logics for describing the behaviour of intelligent agents. We concentrate on a core logic within this framework, in particular, we describe two new methods for providing proof methods for this core logic, discuss some of the problems we have encountered in their design, and present an extended example of the use of the KARO framework and the two proof methods. 1 Agents 537667 Least Squares Conformal Maps for Automatic Texture Atlas Generation A Texture Atlas is an efficient color representation for 3D Paint Systems. The model to be textured is decomposed into charts homeomorphic to discs, each chart is parameterized, and the unfolded charts are packed in texture space. Existing texture atlas methods for triangulated surfaces suffer from several limitations, requiring them to generate a large number of small charts with simple borders. The discontinuities between the charts cause artifacts, and make it difficult to paint large areas with regular patterns. AI dicaro98antnet AntNet: Distributed Stigmergetic Control for Communications Networks This paper introduces AntNet, a novel approach to the adaptive learning of routing tables in communications networks. AntNet is a distributed, mobile agents based Monte Carlo system that was inspired by recent work on the ant colony metaphor for solving optimization problems. AntNet's agents concurrently explore the network and exchange collected information. The communication among the agents is indirect and asynchronous, mediated by the network itself. This form of communication is typical of social insects and is called stigmergy. We compare our algorithm with six state-of-the-art routing algorithms coming from the telecommunications and machine learning elds. The algorithms' performance is evaluated over a set of realistic testbeds. We run many experiments over real and arti cial IP datagram networks with increasing number of nodes and under several paradigmatic spatial and temporal tra c distributions. Results are very encouraging. AntNet showed superior performance under all the experimental conditions with respect to its competitors. We analyze the main characteristics of the algorithm and try to explain the reasons for its superiority. 1. ML rauber01automatically Automatically Analyzing and Organizing Music Archives . We are experiencing a tremendous increase in the amount of music being made available in digital form. With the creation of large multimedia collections, however, we need to devise ways to make those collections accessible to the users. While music repositories exist today, they mostly limit access to their content to query-based retrieval of their items based on textual meta-information, with some advanced systems supporting acoustic queries. What we would like to have additionally, is a way to facilitate exploration of musical libraries. We thus need to automatically organize music according to its sound characteristics in such a way that we nd similar pieces of music grouped together, allowing us to nd a classical section, or a hard-rock section etc. in a music repository. In this paper we present an approach to obtain such an organization of music data based on an extension to our SOMLib digital library system for text documents. Particularly, we employ the Self-Organizing Map to create a map of a musical archive, where pieces of music with similar sound characteristics are organized next to each other on the two-dimensional map display. Locating a piece of music on the map then leaves you with related music next to it, allowing intuitive exploration of a music archive. Keywords: Multimedia, Music Library, Self-Organizing Map (SOM), Exploration of Information Spaces, User Interface, MP3 1 IR yang00improving Improving Text Categorization Methods for Event Tracking Automated tracking of events from chronologically ordered document streams is a new challenge for statistical text classification. Existing learning techniques must be adapted or improved in order to effectively handle difficult situations where the number of positive training instances per event is extremely small, the majority of training documents are unlabelled, and most of the events have a short duration in time. We adapted several supervised text categorization methods, specifically several new variants of the k-Nearest Neighbor (kNN) algorithm and a Rocchio approach, to track events. All of these methods showed significant improvement (up to 71% reduction in weighted error rates) over the performance of the original kNN algorithm on TDT benchmark collections, making kNN among the top-performing systems in the recent TDT3 official evaluation. Furthermore, by combining these methods, we significantly reduced the variance in performance of our event tracking system over different ... IR 32409 The Well-founded Semantics Is the Principle of Inductive Definition . Existing formalisations of (transfinite) inductive definitions in constructive mathematics are reviewed and strong correspondences with LP under least model and perfect model semantics become apparent. I point to fundamental restrictions of these existing formalisations and argue that the well-founded semantics (wfs) overcomes these problems and hence, provides a superior formalisation of the principle of inductive definition. The contribution of this study for LP is that it (re- )introduces the knowledge theoretic interpretation of LP as a logic for representing definitional knowledge. I point to fundamental differences between this knowledge theoretic interpretation of LP and the more commonly known interpretations of LP as default theories or auto-epistemic theories. The relevance is that differences in knowledge theoretic interpretation have strong impact on knowledge representation methodology and on extensions of the LP formalism, for example for representing uncertainty. Keywo... AI 171304 An Empirical Comparison of Decision Trees and Other Classification Methods Twenty two decision tree, nine statistical, and two neural network classifiers are compared on thirtytwo datasets in terms of classification error rate, computational time, and (in the case of trees) number of terminal nodes. It is found that the average error rates for a majority of the classifiers are not statistically significant but the computational times of the classifiers differ over a wide range. The statistical POLYCLASS classifier based on a logistic regression spline algorithm has the lowest average error rate. However, it is also one of the most computationally intensive. The classifier based on standard polytomous logistic regression and a decision tree classifier using the QUEST algorithm with linear splits have the second lowest average error rates and are about 50 times faster than POLYCLASS. Among decision tree classifiers with univariate splits, the classifiers based on the C4.5, IND-CART, and QUEST algorithms have the best combination of error rate and speed, althoug... ML 530008 A Framework for Knowledge Management on the Semantic Web The Semantic Web can be a very promising platform for developing knowledge management systems. However, the problem is how to represent knowledge in the machine-understandable form, so that relevant knowledge can be found by machine agents. In this paper we present a knowledge management approach based on RDF-compatible format for representing rules and on a novel technique for the annotation of knowledge sources by using conditional statements. The approach is based on our existing Semantic Web tools. The main benefit is high improvement in the precision by searching for knowledge, as well as the possibility to retrieve a composition of knowledge sources which are relevant for the problem solving. IR schiele99attentional Attentional Objects for Visual Context Understanding This paper exploits wearable computers' unique opportunity to record and index the visual environment of the user from the "first-person" perspective. We propose to use a hat-mounted wearable camera to record what the user sees during the day with a wearable computer. This camera can be used to make the computer more contextually aware of the user and their actions. Furthermore, the camera can be used to record, analyze and index the visual environment of the user. By keeping track of the actions of the user upon and within the environment the system can be more aware of the interactions of the user within the environment. An important aspect of the system is to automatically extract objects of user interest, and their motion within the environment and relative to the user. 1 Introduction Wearable computers have the potential to "see" as the user sees, "hear" as the user hears, and experience the life and the environment of the user in a "first-person" sense. As has been pointed out ... HCI 445758 Extending a Multi-Agent System for Genomic Annotation . The explosive growth in genomic (and soon, expression and proteomic) data, exemplified by the Human Genome Project, is a fertile domain for the application of multi-agent information gathering technologies. Furthermore, hundreds of smaller-profile, yet still economically important organisms are being studied that require the efficient and inexpensive automated analysis tools that multiagent approaches can provide. In this paper we give a progress report on the use of the DECAF multi-agent toolkit to build reusable information gathering systems for bioinformatics. We will briefly summarize why bioinformatics is a classic application for information gathering, how DECAF supports it, and recent extensions underway to support new analysis paths for genomic information. 1 Agents wolski98fuzzy Fuzzy Triggers: Incorporating Imprecise Reasoning into Active Databases Traditional Event-Condition-Action triggers (active database rules) include a Boolean predicate as a trigger condition. We propose fuzzy triggers whereby fuzzy inference is utilized in the condition evaluation. This way, approximate reasoning may be integrated with a traditional crisp database. The new approach paves the way for intuitive expression of application semantics of imprecise nature, in database-bound applications. Two fuzzy triggers models are proposed. Firstly, a set of fuzzy rules is encapsulated into a Boolean-valued function called a rule set function, leading to the C-fuzzy trigger model. Subsequently, actions are expressed also in fuzzy terms, and the corresponding CA-fuzzy trigger model is proposed. Examples are provided to illustrate how fuzzy triggers can be applied to a real-life drive control system in an industrial installation. 1 . Introduction There has been considerable interest in active database rules (called triggers in commercial applications, and in th... ML chen00websail WebSail: From On-line Learning to Web Search In this paper we investigate the applicability of on-line learning algorithms to the real-world problem of web search. Consider that web documents are indexed using n Boolean features. We first present a practically efficient on-line learning algorithm TW2 to search for web documents represented by a disjunction of at most k relevant features. We then design and implement WebSail, a real-time adaptive web search learner, with TW2 as its learning component. WebSail learns from the user's relevance feedback in real-time and helps the user to search for the desired web documents. The architecture and performance of WebSail are also discussed. IR pinheirodasilva00user User Interface Modelling with UML The Unified Modeling Language (UML) is a natural candidate for user interface (UI) modelling since it is the standard notation for object oriented modelling of applications. However, it is by no means clear how to model UIs using UML. This paper presents a user interface modelling case study using UML. This case study identifies some aspects of UIs that cannot be modelled using UML notation, and a set of UML constructors that may be used to model UIs. The modelling problems indicate some weaknesses of UML for modelling UIs, while the constructors exploited indicate some strengths. The identification of such strengths and weaknesses can be used in the formulation of a strategy for extending UML to provide greater support for user interface design. HCI klesen00exploiting Exploiting Models of Personality and Emotions to Control the Behavior of Animated Interactive Agents The German Research Centre for Artificial Intelligence (DFKI) recently started three new projects1 to advance our understanding of the fundamental technology required to drive the social behaviour of interactive animated agents. This initiative has been timed to catch the current wave of research and commercial interest in the field of lifelike characters [1] and HCI 289242 Offering a Precision-Performance Tradeoff for Aggregation Queries over Replicated Data Strict consistency of replicated data is infeasible or not required by many distributed applications, so current systems often permit stale replication,inwhich cached copies of data values are allowed to become out of date. Queries over cached data return an answer quickly, but the stale answer may be unboundedly imprecise. Alternatively, queries over remote master data return a precise answer, but with potentially poor performance. To bridge the gap between these two extremes, we propose a new class of replication systems called TRAPP (Tradeoff in Replication Precision and Performance). TRAPP systems give each user fine-grained control over the tradeoff between precision and performance: Caches store ranges that are guaranteed to bound the current data values, instead of storing stale exact values. Users supply a quantitative precision constraint along with each query. To answer a query, TRAPP systems automatically select a combination of locally cached bounds and exact master data stored remotely to deliver a bounded answer consisting of a range that is no wider than the specified precision constraint, that is guaranteed to contain the precise answer, and that is computed as quickly as possible. This paper defines the architecture of TRAPP replication systems and covers some mechanics of caching data ranges. It then focuses on queries with aggregation, presenting optimization algorithms for answering queries with precision constraints, and reporting on performance experiments that demonstrate the fine-grained control of the precision-performance tradeoff offered by TRAPP systems. DB greenberg98using Using a Room Metaphor to Ease Transitions in Groupware Many groupware systems contain gaps that hinder or block natural social interaction or that do not let people easily move between different styles of work. We believe that the adoption of a room metaphor can ease people's transitions across these gaps, allowing them to work together more naturally. Using the TeamWave Workplace system as an example, we show how particular gaps are removed. First, we ease a person's transition between single user and groupware applications by making rooms suitable for both individual and group activity. Second, people can move fluidly between asynchronous and synchronous work because room artifacts persist. People can leave messages, documents and annotations for others, or work on them together when occupying the room at the same time. Third, we ease the difficulty of initiating real time work by providing people with awareness of others who may be available for real-time interactions, and by automatically establishing connections as users enter a commo... HCI dotsch00tic TIC - A Toolkit for Validation in Formal Language Learning Quite often, heuristics and common sense suggest directions for improving well--known learning algorithms. However it seems not an easy task to verify that the modifications are indeed helpful. This is made more complicated through various additional influences inherent in different application domains. In order to obtain a faithful impression of phenomena that are intrinsic to the algorithms, the role of specific domains should be minimized. Our validation toolkit TIC allows to explore the behaviour of various algorithms for learning formal languages. This is a well-examined and standardized application domain. TIC is operated by interactive as well as automatic control. Motivation and Introduction Today, a lot of different learning approaches and algorithms do exist. There are "classical" as well as "brand new" approaches, and all of them come in many versions and refinements. On the one hand this indicates a desirable improvement of methods, but on the other hand it ... ML goller99connectionist A Connectionist Approach for Learning Search-Control Heuristics for Automated Deduction Systems The central problem in automated deduction is the explosive growth of search spaces when proof length increases. In this paper, a connectionist approach for learning search-control heuristics for automated deduction systems is presented. In particular, we show how folding architecture networks, a new type of neural networks capable of solving supervised learning tasks on structured data, can be used for learning heuristics evaluation functions for algebraic (logical) expressions and how these evaluation functions can then be used to control the search process for new proof problems. Experimental results with the automated deduction system Setheo in an algebraic domain show a considerable performance improvement. Controlled by heuristics which had been learned from simple problems in this domain the system is able to solve several problems from the same domain which had been out of reach for the original system. 1 Introduction The goal in automated deduction (AD) is to automatically... ML cremonini00ruling Ruling Agent Motion in Structured Environments . The design and development of cooperative Internet applications based on mobile agents require appropriate modelling of both the physical space where agents roam and the conceptual space of mobile agent interaction. The paper discusses how an open, Internet-based, organisation network can be modelled as a hierarchical collection of locality domains, where agents can dynamically acquire information about resource location and availability according to their permissions. It also analyses the issue of how agent motion can be ruled and constrained within a structured environment by means of an appropriate coordination infrastructure. 1 Introduction Mobile agents are a promising technology for the design and development of cooperative applications on the Internet [3, 5, 12, 13]. Due to their capability of autonomously roaming the Internet, mobile agents can move locally to the resources they need -- let them be users, data, or services -- and there interact with them. This can p... Agents jain99statistical Statistical Pattern Recognition: A Review AbstractÐThe primary goal of pattern recognition is supervised or unsupervised classification. Among the various frameworks in which pattern recognition has been traditionally formulated, the statistical approach has been most intensively studied and used in practice. More recently, neural network techniques and methods imported from statistical learning theory have been receiving increasing attention. The design of a recognition system requires careful attention to the following issues: definition of pattern classes, sensing environment, pattern representation, feature extraction and selection, cluster analysis, classifier design and learning, selection of training and test samples, and performance evaluation. In spite of almost 50 years of research and development in this field, the general problem of recognizing complex patterns with arbitrary orientation, location, and scale remains unsolved. New and emerging applications, such as data mining, web searching, retrieval of multimedia data, face recognition, and cursive handwriting recognition, require robust and efficient pattern recognition techniques. The objective of this review paper is to summarize and compare some of the well-known methods used in various stages of a pattern recognition system and identify research topics and applications which are at the forefront of this exciting and challenging field. ML amiri00dynamic Dynamic Function Placement for Data-intensive Cluster Computing Optimally partitioning application andfilesystem functionality within a cluster of clients and servers is a difficult problem dueto dynamic variations in application behavior, resource availability, and workload mixes. Thispaper presents A BACUS, a run-time systemthat monitors and dynamically changes function placement for applications that manipulate largedata sets. Several examples of data-intensive workloads are used to show the importance ofproper function placement and its dependence on dynamic run-time characteristics, withperformance differences frequently reaching 2-10X. We evaluate how well the ABACUSprototype adapts to run-time system behavior, including both long-term variation (e.g., filterselectivity) and short-term variation (e.g., multi-phase applications and inter-applicationresource contention). Our experiments with ABACUS indicate that it is possible to adapt inall of these situations and that the adaptation converges most quickly in those cases where theperformance impact is most significant. 1 DB 297471 Subsumption for XML Types . XML data is often used (validated, stored, queried, etc) with respect to different types. Understanding the relationship between these types can provide important information for manipulating this data. We propose a notion of subsumption for XML to capture such relationships. Subsumption relies on a syntactic mapping between types, and can be used for facilitating validation and query processing. We study the properties of subsumption, in particular the notion of the greatest lower bound of two schemas, and show how this can be used as a guide for selecting a storage structure. While less powerful than inclusion, subsumption generalizes several other mechanisms for reusing types, notably extension and refinement from XML Schema, and subtyping. 1 Introduction XML [5] is a data format for Web applications. As opposed to e.g., relational databases, XML documents do not have to be created and used with respect to a fixed, existing schema. This is particularly useful in Web ap... DB lerman00coalition Coalition Formation for Large-Scale Electronic Markets Coalition formation is a desirable behavior in a multiagent system, when a group of agents can perform a task more efficiently than any single agent can. Computational and communications complexity of traditional approaches to coalition formation, e.g., through negotiation, make them impractical for large systems. We propose an alternative, physics-motivated mechanism for coalition formation that treats agents as randomly moving, locally interacting entities. A new coalition may form when two agents encounter one another, and it may grow when a single agent encounters it. Such agent-level behavior leads to a macroscopic model that describes how the number and distribution of coalitions change with time. We increase the generality and complexity of the model by letting the agents leave coalitions with some probability. The model is expressed mathematically as a series of differential equations. These equations have steady state solutions that describe the equilibrium distribution of coa... Agents dolan00benchmarking Benchmarking Optimization Software with COPS 1 Introduction 1 Testing Methods 2 1 Largest Small Polygon 3 2 Distribution of Electrons on a Sphere 5 3 Hanging Chain 7 4 Shape Optimization of a Cam 9 5 Isometrization of ff-pinene 11 6 Marine Population Dynamics 13 7 Flow in a Channel 16 8 Robot Arm 18 9 Particle Steering 21 10 Goddard Rocket 23 11 Hang Glider 26 12 Catalytic Cracking of Gas Oil 29 13 Methanol to Hydrocarbons 31 14 Catalyst Mixing 33 15 Elastic-Plastic Torsion 35 16 Journal Bearing 37 17 Minimal Surface with Obstacle 39 Acknowledgments 41 References 41 ii Benchmarking Optimization Software with COPS by Elizabeth D. Dolan and Jorge J. Mor'e Abstract We describe version 2.0 of the COPS set of nonlinearly constrained optimization problems. We have added new problems, as well as streamlined and improved most of the problems. We also provide a comparison of the LANCELOT, LOQO, MINOS, and SNOPT solvers on these problems. Introduction The COPS [5] test set provides a modest selection of difficult nonlinearly constrai... AI sturm00tableau A Tableau Calculus for Temporal Description Logic: The Expanding Domain Case . In this paper we present a tableau calculus for a temporal extension of the description logic ALC, called T LALC . This logic is based on the temporal language with `Until' interpreted over the natural numbers with expanding ALC-domains. The tableau calculus forms an elaborated combination of Wolper's tableau calculus for propositional linear temporal logic, the standard tableau-algorithm for ALC, and the method of quasimodels as it has been introduced by Wolter and Zakharyaschev. Based on those three ingredients the paper provides a new method of how tableau-based decision procedures can be constructed for many-dimensional logics which lack the finite model property. The method can be applied to deal with other temporalized formalisms as well. 1 Introduction In many application domains of logic in Computer Science and Artificial Intelligence it is no longer enough to describe the static aspect of the world. In particular, there is a need to formalize its temporal evolution... DB lesh99using Using Plan Recognition in Human-Computer Collaboration . Human-computer collaboration provides a practical and useful application for plan recognition techniques. We describe a plan recognition algorithm which is tractable by virtue of exploiting properties of the collaborative setting, namely: the focus of attention, the use of partially elaborated hierarchical plans, and the possibility of asking for clarification. We demonstrate how the addition of our plan recognition algorithm to an implemented collaborative system reduces the amount of communication required from the user. 1 Introduction An important trend in recent work on human-computer interaction and user modeling has been to view human-computer interaction as a kind of collaboration (e.g, Ferguson and Allen, 1998, Guinn, 1996, Rich and Sidner, 1998, Rickel and Johnson, 1998). In this approach, the human user and the computer (often personified as an "agent") coordinate their actions toward achieving shared goals. A common setting for collaboration, illustrated in Figure 1... HCI 5188 Empirical Risk Approximation: An Induction Principle for Unsupervised Learning Unsupervised learning algorithms are designed to extract structure from data without reference to explicit teacher information. The quality of the learned structure is determined by a cost function which guides the learning process. This paper proposes Empirical Risk Approximation as a new induction principle for unsupervised learning. The complexity of the unsupervised learning models are automatically controlled by the two conditions for learning: (i) the empirical risk of learning should uniformly converge towards the expected risk; (ii) the hypothesis class should retain a minimal variety for consistent inference. The maximal entropy principle with deterministic annealing as an efficient search strategy arises from the Empirical Risk Approximation principle as the optimal inference strategy for large learning problems. Parameter selection of learnable data structures is demonstrated for the case of k-means clustering. 1 What is unsupervised learning? Learning algorithms are desi... ML simmons00first First Results in the Coordination of Heterogeneous Robots for Large-Scale Assembly : While many multi-robot systems rely on fortuitous cooperation between agents, some tasks, such as the assembly of large structures, require tighter coordination. We present a general software architecture for coordinating heterogeneous robots that allows for both autonomy of the individual agents as well as explicit coordination. This paper presents recent results with three robots with very different configurations. Working as a team, these robots are able to perform a high-precision docking task that none could achieve individually. 1. Introduction As robots become more autonomous and sophisticated, they are increasingly being used for more complex and demanding tasks. Often, single robots are insufficient to perform the tasks. For some types of tasks, such as exploration or demining, multiple robots can be used to increase efficiency and reliability. For many other tasks, however, not only are multiple robots necessary, but explicit coordination amongst the robots is imper... Agents 9683 A Control Architecture for Flexible Internet Auction Servers The flexibility to support both high activity and low activity auctions is required by any system that allows bidding by both humans and software agents. We present the control architecture of the Michigan Internet AuctionBot, and discuss some of the system engineering issues that arose in its design. 1 Introduction The Michigan Internet AuctionBot is a highly configurable auction server built to support research on electronic commerce and multiagent negotiation [3]. The first generation architecture was simple and robust, and allowed us to concentrate on other aspects of the system. However, several inefficiencies made it problematic to run auctions with very fast interactions. We have redesigned the core AuctionBot architecture in order to improve overall performance, while still meeting the original goal: a system that is configurable, maintainable, and capable of conducting a large number of simultaneous auctions. In AuctionBot architecture nomenclature, we say an auction is open ... DB 534081 Active Proxy-G: Optimizing the Query Execution Process in the Grid The Grid environment facilitates collaborative work and allows many users to query and process data over geographically dispersed data repositories. Over the past several years, there has been a growing interest in developing applications that interactively analyze datasets, potentially in a collaborative setting. We describe an Active Proxy-G service that is able to cache query results, use those results for answering new incoming queries, generate subqueries for the parts of a query that cannot be produced from the cache, and submit the subqueries for final processing at application servers that store the raw datasets. We present an experimental evaluation to illustrate the effects of various design tradeoJj5 . We also show the benefits that two real applications gain from using the middleware. IR 510049 A Semiotic Communication Model for Interface Design This research wants to contribute to the creation of a semiotic framework for interface design. Using the Jakobson's communication model to analyse the HCI approach to interface development, we explain how some central factors of communication are not enough considered by designers. HCI 103027 Modeling Emotions and Other Motivations in Synthetic Agents We present Cathexis, a distributed, computational model which offers an alternative approach to model the dynamic nature of different affective phenomena, such as emotions, moods and temperaments, and provides a flexible way of modeling their influence on the behavior of synthetic autonomous agents. The model has been implemented as part of an extensible, object-oriented framework which provides enough functionality for agent developers to design emotional agents that can be used in a variety of applications including entertainment (e.g. synthetic agents for interactive drama, video games, etc.), education (e.g. Intelligent Tutoring Systems), and human-computer interfaces. Introduction Emotions are an essential part of our lives, they influence how we think and behave, and how we communicate with others. Several researchers have acknowledged their importance in human thinking [Minsky 1986; Toda 1993], and recent neurological evidence seems to support these ideas [LeDoux 1996; Damasio... Agents srinivasan02web Web Crawling Agents for Retrieving Biomedical Information Autonomous agents for topic driven retrieval of information from the Web are currently a very active area of research. The ability to conduct real time searches for information is important for many users including biomedical scientists, health care professionals and the general public. We present preliminary research on different retrieval agents tested on their ability to retrieve biomedical information, whose relevance is assessed using both genetic and ontological expertise. In particular, the agents are judged on their performance in fetching information about diseases when given information about genes. We discuss several key insights into the particular challenges of agent based retrieval learned from our initial experience in the biomedical domain. IR fan01splitting On splitting and Cloning Agents Embedded with cloning mechanisms, an agent can balance its own loads by discharging computing tasks to its clones when it is over-loaded. In addition, it's more reasonable to transfer the smarter, smaller clones of an agent rather than the bulky agent itself in mobile computing. In this paper, a simple BDI agent model is formally established. Using this model, the semantics of constructing new agents by inheritance and self-identifying behavior of existing agents are precisely de ned. Four kinds of cloning mechanisms are identi ed, the properties of each cloning mechanism and the relationships in between are studied, and some implementation issues are also discussed. Agents baumgartner01supervised Supervised Wrapper Generation with Lixto We illustrate basic features of the Lixto wrapper generator such as the user and system interaction, the capacious visual interface, the marking and selecting procedures, and the extraction tasks by describing the construction of a simple example program in the current Lixto prototype. 1. IR carbonell98report Report on the CONALD Workshop on Learning from Text and the Web Moo], organization and presentation of documents in information retrieval systems [GS, Hof], collaborative filtering [dVN], lexicon learning [GBGH], query reformulation [KK], text generation [Rad] and analysis of the statistical properties of text [MA]. In short, the state of the art in learning from text and the web is that a broad range of methods are currently being applied to many important and interesting tasks. There remain numerous open research questions, however. Broadly, the goals of the work presented at the workshop fall into two overlapping categories: (i) making textual information available in a structured format so that it can be used for complex queries and problem solving, and (ii) assisting users in finding, organizing and managing information represented in text sources. As an example of research aimed at the former goal, Muslea, Minton and Knoblock [MMK] have developed an approach to learning wrappers for semi-structured Web sources, such as restau IR 515869 A Two-stage Scheme for Dynamic Hand Gesture Recognition In this paper a scheme is presented for recognizing hand gestures using the output of a hand tracker which tracks a rectangular window bounding the hand region. A hierarchical scheme for dynamic hand gesture recognition is proposed based on state representation of the dominant feature trajectories using an a priori knowledge of the way in which each gesture is performed. HCI coppin00eventscope EventScope: Amplifying Human Knowledge and Experience via Intelligent Robotic Systems and Information Interaction The EventScope program develops publicly accessible "reality browsers" that display both archived and updating representations of remote environments derived from on-site robotic sensors. The interface encourages collaborative work within a community of users. Public exploration of real remote sites presents a variety of interface issues addressed by EventScope, including time delay, public exploration via a single robot and communication between geographically separate users from diverse backgrounds. Merging public interface with educational and contextual information extends the notion of "interface" to "remote reality library." EventScope is a NASA and private foundationfunded project based at Carnegie Mellon University. 1. Introduction Publicly funded Earth and planetary exploration is conducted to increase knowledge of our universe. The public traditionally accesses this knowledge passively, through the media. However, the development of the Web and of robotic remote-sensing tech... HCI bertelsen99dynamics Dynamics in Wastewater Treatment: A Framework for Understanding Formal Constructs in Complex Technical Settings . Based on the study of unskilled work in a Danish wastewater treatment plant, the problem of formalisation of work is discussed and extended to technical processes. Five symmetrical levels of dynamics in complex technical work arrangements are proposed as a tool for understanding the limits of formalisation and for designing formal constructs in such settings. The analysis is based on concepts of heterogeneity, granularity of goals and motives, and process and structure. Introduction An inevitable problem in the design of CSCW systems is that work is not standing still. Rather, work settings are dynamic: routines evolve over time and unusual situations force deviations from the routine. For this reason, purely formal constructs and descriptions of work have proven inadequate when designing effective, real world CSCW systems. Suchman and Wynn (1984) set off the debate about the role of formalism in CSCW, and their empirical studies clearly illustrated that there is more to office wo... HCI warshaw98monitoring Monitoring Network Logs for Anomalous Activity We report on the progress of the VenusDB active-database system as driven by WatchDog, an application in network intrusion detection. The application is typical of a class of problems we coin monotonic log monitoring systems. These are systems where real-time data sources are logged to a database for transactional assurances and the database further provides services for decision support. Milestones comprise the successful layering of a Venus language executable with Oracle through the use of the Venus Abstract Machine Interface (AMI, a data abstraction interface) and Oracle's native trigger mechanism. The identification of monotonic-logging systems as an interesting application class enables us to limit coupling modes and to identify an effective layered architecture. DB davison00topical Topical Locality in the Web: Experiments and Observations Most web pages are linked to others with related content. This idea, combined with another that says that text in, and possibly around, HTML anchors describe the pages to which they point, is the foundation for a usable World-Wide Web. In this paper, we examine to what extent these ideas hold by empirically testing whether topical locality mirrors spatial locality of pages on the Web. In particular, we find that the likelihood of linked pages having similar textual content to be high; the similarity of sibling pages increases when the links from the parent are close together; titles, descriptions, and anchor text represent at least part of the target page; and that anchor text may be a useful discriminator among unseen child pages. These results present the foundations necessary for the success of many web systems, including search engines, focused crawlers, linkage analyzers, and intelligent web agents. 1 Introduction Most web pages are linked to others with related content... IR kollios99nearest Nearest Neighbor Queries in a Mobile Environment Nearest neighbor queries have received much interest in recent years due to their increased importance in advanced database applications. However, past work DB bassiliades00edevice E-DEVICE: An Extensible Active Knowledge Base System with Multiple Rule Type Support This paper describes E-DEVICE, an extensible active knowledge base system (KBS) that supports the processing of event-driven, production, and deductive rules into the same active OODB system. E-DEVICE provides the infrastructure for the smooth integration of various declarative rule types, such as production and deductive rules, into an active OODB system that supports low-level event-driven rules only by a) mapping each declarative rule into one event-driven rule, offering centralized rule selection control for correct run-time behavior and conflict resolution, and b) using complex events to map the conditions of declarative rules and monitor the database to incrementally match those conditions. E-DEVICE provides the infrastructure for easily extending the system by adding a) new rule types as subtypes of existing ones and b) transparent optimizations to the rule matching network. The resulting system is a flexible, yet efficient, KBS that gives the user the ability to express knowledge in a variety of high-level forms for advanced problem solving in data intensive applications. DB camerer00ewa EWA Learning in Bilateral Call Markets This chapter extends the EWA learning model to bilateral call market games (also known as the "sealed-bid mechanism" in two-person bargaining). In these games, a buyer and seller independently draw private values from commonly-known distributions and submit bids. If the buyer's bid is above the seller's, they trade at the midpoint of the two bids; otherwise they don't trade. We apply EWA by assuming that players have value-dependent bidding strategies, and they partially generalize experience from one value/cost condition to another in response to the incentives from nonlinear optimal bid functions. The same learning model can be applied to other market institutions where subjects economize on learning by taking into consideration similarity between past experience and a new environment while still recognizing the difference in market incentives between them. The chapter also presents a new application of EWA to a "continental divide" coordination game, and reviews 32 earlier studies comparing EWA, reinforcement, and belief learning. The application shows the advantages of a generalized adaptive model of behavior that includes elements of reinforcement, belief-based and direction learning as special cases at some cost of complexity for the benefit of generality and psychological appeal. It is a good foundation to build upon to extend our understanding of adaptive behavior in more general games and market institutions. In future work, we should investigate the similarity parameters, y and w, to better characterize their magnitude and significance in different market institutions. Keywords: Experimental economics, call markets, sealed-bid mechanism, learning JEL Classification: D44, D83, C92 August 2, 2000. Thanks to Terry Daniel for supplying data. This research has been... ML koch99simple A Simple Query Facility for the Objectivity/DB Persistent Object Manager This document discusses the reasons that lead to the development of a query faciliy within the CRISTAL project, its design criteria, syntax, semantics, use, and restrictions. It is furthermore intended to serve as a preliminary manual. The query facility is discussed in its immediately next development stage which should be finished within the next few weeks DB crabbe01multiple Multiple Goal Q-Learning: Issues and Functions This paper addresses the concerns of agents using reinforcement learning to learn to achieve multiple simultaneous goals. It proves that an algorithm based on acting upon the maximal goal at any one time will, in many cases, not not produce the Maximal Expected Utility for the agent. The paper then examines the type of function approximator necessary for the agent's reinforcement learning system, and concludes that a bi-linear function is the best compromise between expressive power and speed of learning. ML flake02extracting Extracting Query Modifications from Nonlinear SVMs When searching the WWW, users often desire results restricted to a particular document category. Ideally, a user would be able to filter results with a text classifier to minimize false positive results; however, current search engines allow only simple query modifications. To automate the process of generating effective query modifications, we introduce a sensitivity analysis-based method for extracting rules from nonlinear support vector machines. The proposed method allows the user to specify a desired precision while attempting to maximize the recall. Our method performs several levels of dimensionality reduction and is vastly faster than searching the combination feature space; moreover, it is very effective on real-world data. IR 142848 Extending the ODMG Object Model with Time Although many temporal extensions of the relational data model have been proposed, there is no comparable amount of work in the context of object-oriented data models. Moreover, extensions to the relational model have been proposed in the framework of SQL standards, whereas no attempts have been made to extend the standard for object-oriented databases, defined by ODMG. This paper presents T ODMG, a temporal extension of the ODMG-93 standard data model. The main contributions of this work are, thus, the formalization of the ODMG standard data model and its extension with time. Another contribution of this work is the investigation, on a formal basis, of the main issues arising from the introduction of time in an object-oriented model. DB aiello01ontological Ontological Overhearing The collaboration between two intelligent agents can be greatly enhanced if a third agent, who has some understanding of the communication between the first two, intervenes giving appropriate information or acting helpfully without having been explicitly involved. The behavior of this third agent, quite common in human interaction, is called overhearing. We present an agent architecture modeling this behavior. In particular, we focus on overhearing based on ontological reasoning; that is, the overhearer semantically selects pieces of communication according to his own knowledge (ontologically organized) and goals. Agents schweighofer01improving Improving the Quality of Labels for Self-Organising Maps Using Fine-Tuning Vector representation of legal documents is still the best way for computing classification clusters and labelling of its contents. A very special problem occurs with self organising maps: strong clusters tend to dominate neighbouring smaller clusters in terms of their weight vector structure, which influences the labels extracted from these. This unwelcome side-effect can be overcome efficiently with a dedicated fine-tuning phase at the end of the training process, in which the neighbourhood radius of the training function is set to zero. Experiments with our text collection have shown the high improvement of the quality of labelling. IR gaskett00reinforcement Reinforcement Learning for Visual Servoing of a Mobile Robot A novel reinforcement learning algorithm is applied to a visual servoing task on a real mobile robot. There is no requirement for camera calibration, an actuator model or a knowledgeable teacher. The controller learns from a critic which gives a scalar reward. The learning algorithm handles continuously valued states and actions and can learn from good and bad experiences including data gathered while performing unrelated behaviours and from historical data. Experimental results are presented. 1 Introduction Visual servoing consists of moving some part of a robot to a desired position using visual feedback [ Hutchinson et al., 1996 ] . It is a basic building block for purposeful robot behaviours such as foraging, target pursuit and landmark based navigation. Some degree of calibration is generally required to achieve visual servoing. This calibration can be a time consuming and error prone process. In this work we show that reinforcement based learning can eliminate the ca... ML duch00optimization Optimization and Interpretation of Rule-Based Classifiers Abstract. Machine learning methods are frequently used to create rule-based classifiers. For continuous features linguistic variables used in conditions of the rules are defined by membership functions. These linguistic variables should be optimized at the level of single rules or sets of rules. Assuming the Gaussian uncertainty of input values allows to increase the accuracy of predictions and to estimate probabilities of different classes. Detailed interpretation of relevant rules is possible using (probabilistic) confidence intervals. A real life example of such interpretation is given for personality disorders. The approach to optimization and interpretation described here is applicable to any rule-based system. 1 Introduction. In many applications rule-based classifiers are created starting from machine learning, fuzzy logic or neural network methods [1]–[3]. If the number of rules is relatively small and accuracy is sufficiently high such classifiers are an optimal choice, because the reasons for their decisions are easily verified. Crisp logical rules are desirable ML ester00dctree The DC-tree: A Fully Dynamic Index Structure for Data Warehouses : Many companies have recognized the strategic importance of the knowledge hidden in their large databases and have built data warehouses. Typically, updates are collected and applied to the data warehouse periodically in a batch mode, e.g., over night. Then, all derived information such as index structures has to be updated as well. The standard approach of bulk incremental updates to data warehouses has some drawbacks.First, the average runtime for a single update is small but the total runtime for the whole batch of updates may become rather large. Second, the contents of the data warehouse is not always up to date. In this paper, we introduced the DC-tree, a fully dynamic index structure for data warehouses modeled as a data cube. This new index structure is designed for applications where the above drawbacks of the bulk update approach are critical. The DC-tree is a hierarchical index structure - similar to the X-tree - exploiting the concept hierarchies typically defined for the... DB chen99introducing Introducing a New Advantage of Crossover: Commonality-Based Selection The Commonality-Based Crossover Framework defines crossover as a two-step process: 1) preserve the maximal common schema of two parents, and 2) complete the solution with a construction heuristic. In these “heuristic ” operators, the first step is a form of selection. This commonality-based form of selection has been isolated in GENIE. Using random parent selection and a non-elitist generational replacement scheme, GENIE does not include fitness-based selection. However, a theoretical analysis shows that “ideal ” construction heuristics in GENIE can potentially converge to optimal solutions. Experimentally, results show that the effectiveness of practical construction heuristics can be amplified by commonalitybased restarts. Overall, it is shown that the commonality hypothesis is valid--schemata common to above-average solutions are indeed above average. Since common schemata can only be identified by multi-parent operators, commonality-based selection is a unique advantage that crossover can enjoy over mutation. 1 ML kon00flexible A Flexible, Interoperable Framework for Active Spaces this paper we describe the requirements faced by such a system and propose an integrated architecture meeting these requirements. The paper focuses on a representation of Active Spaces using standard Naming and Trading mechanisms and on an object-oriented framework for managing heterogeneous devices. HCI mcdonald00heterogeneous Heterogeneous Database Integration Using Agent-Oriented Information Systems : The Department of Defense (DOD) has an extensive family of models used to simulate the mission level interaction of weapon systems. Interoperability and reuse of the underlying data files used to create simulation scenarios pose great challenges in this regard. Unlike traditional data integration methods common to federated database research, the emerging field of agent-oriented information systems (AOIS) views data as the central focus of an application while also providing an overall architectural framework for application development. We develop an AOIS solution relevant to this problem domain by combining object-oriented data modeling (OMT), a persistent programming language using a commercial objectoriented database (ObjectStore#), and an agentoriented analysis and design methodology (MaSE). Requirements from a contractor-led effort at the Air Force Research Laboratory (AFRL) known as CERTCORT are the basis for analysis and design of our system. We implement prototypical information-layer applications to conceptually demonstrate the reusability and integration of scenarios across simulation models. Keywords: AOIS, Agents, Modeling and Simulations, Heterogeneous Database Integration 1. DB vanschooten01structuring Structuring Distributed Virtual Environments Using a Relational Database Model This paper discusses a specification technique that is based on a traditional (entity-relationship) database model to model the architecture of complex interactive systems, in particular multimodal and multi-user user interfaces. User interface components and other software components ... HCI 455752 Chart of Darkness: Mapping a Large Intranet We introduce and de ne the concept of dark matter on the Web. Dark matter for a person or Web crawler consists of pages that they cannot reach and view, but which another observer can. Dark matter is important to our understanding of the Web in that the portion of the Web any of us can see depends on our viewpoint. Diffrent observers see different overlapping sections of the Web. However, no one can see all of the Web, even if they want to. We categorise the various types of dark matter that exist and how they may be discovered. Formal definitions of what constitutes lightness and darkness on the Web are formulated in terms of reachability. Our case study of dark matter within the Australian National University's intranet is reported. We estimate that 87% of the ANU intranet's information is dark to our local search service, and 37% is potentially loadable Web data unreachable to almost every Web user. Finally, we discuss some of the implications of dark matter for estimating the size of the Web and for general Web searching. IR huffman94instructable Instructable Autonomous Agents INSTRUCTABLE AUTONOMOUS AGENTS by Scott Bradley Huffman Chair: John E. Laird In contrast to current intelligent systems, which must be laboriously programmed for each task they are meant to perform, instructable agents can be taught new tasks and associated knowledge. This thesis presents a general theory of learning from tutorial instruction and its use to produce an instructable agent. Tutorial instruction is a particularly powerful form of instruction, because it allows the instructor to communicate whatever kind of knowledge a student needs at whatever point it is needed. To exploit this broad flexibility, however, a tutorable agent must support a full range of interaction with its instructor to learn a full range of knowledge. Thus, unlike most machine learning tasks, which target deep learning of a single kind of knowledge from a single kind of input, tutorability requires a breadth of learning from a broad range of instructional interactions. The theory of learning from tutorial... ML cohn01qualitative Qualitative Spatial Representation and Reasoning: An Overview . The paper is a overview of the major qualitative spatial representation and reasoning techniques. We survey the main aspects of the representation of qualitative knowledge including ontological aspects, topology, distance, orientation and shape. We also consider qualitative spatial reasoning including reasoning about spatial change. Finally there is a discussion of theoretical results and a glimpse of future work. The paper is a revised and condensed version of [33, 34]. Keywords: Qualitative Spatial Reasoning, Ontology. The text is in a slightly di erent format from the FI format. Cohn HCI 131492 Intelligent Gradient-Based Search of Incompletely Defined Design Spaces Gradient-based numerical optimization of complex engineering designs offers the promise of rapidly producing better designs. However, such methods generally assume that the objective function and constraint functions are continuous, smooth, and defined everywhere. Unfortunately, realistic simulators tend to violate these assumptions. We present a rule-based technique for intelligently computing gradients in the presence of such pathologies in the simulators, and show how this gradient computation method can be used as part of a gradient-based numerical optimization system. We tested the resulting system in the domain of conceptual design of supersonic transport aircraft, and found that using rule-based gradients can decrease the cost of design space search by one or more orders of magnitude. Keywords: Optimization, gradients, sequential quadratic programming, rule-based systems. 1 Introduction Automated search of a space of candidate designs seems an attractive way to improve the tr... ML 249151 CAT: the Copying Approach to Tabling The SLG-WAM is an abstract machine that can be characterized as a sharing approach to implementing tabling: The execution environments of suspended computations are interspersed in the WAM stacks. Stacks are frozen using a set of freeze registers, and the WAM trail mechanism is extended so that the suspended computations can be resumed. This technique has a reasonably small execution overhead, but it is not easy to implement on top of an existing Prolog system. It is also quite difficult to understand. We propose a new technique for the implementation of tabling: the copying approach to tabling. CAT does not impose any overhead to the execution of Prolog code and can be introduced into an existing Prolog system orthogonally. Also, CAT is easier to understand. We have implemented CAT in the XSB system by taking out SLG-WAM and adding CAT. We describe the additions needed for adopting CAT in a WAM implementation. We show a case in which CAT performs arbitrarily worse than SLG-WAM, but on the other hand we present empirical evidence that CAT is competitive and often faster than the SLG-WAM. We also briefly discuss issues related to memory management and to the scheduling. DB yang01pms PMS: a PVC Management System for ATM Networks . Reported in this paper is the developed PMS, a PVC management system for ATM networks. PMS provides a scalable, end-to-end path management solution required for managing todays complex ATM networks. It aims to assist the network operators to perform PVC operations with simplified procedures and automatic optimum route selection. It also aims to provide effective decision-making support for PVC fault identification and prevention to the network operators. 1 Introduction ATM communication network is playing more and more important role in todays telecommunication networks. It has been widely used in backbone networks, transmission networks, access networks, and even enterprise networks. Such emerging large heterogeneous ATM networks have raised many new challenges for researchers and developers in the area of network management. In the management of ATM communication networks that have increased dramatically in size and complexity, the PVC (Permanent Virtual Circuit) managemen... Agents 66208 An Efficient Boosting Algorithm for Combining Preferences We study the problem of learning to accurately rank a set of objects by combining a given collection of ranking or preference functions. This problem of combining preferences arises in several applications, such as that of combining the results of different search engines, or the "collaborativefiltering " problem of ranking movies for a user based on the movie rankings provided by other users. In this work, we begin by presenting a formal framework for this general problem. We then describe and analyze an efficient algorithm called RankBoost for combining preferences based on the boosting approach to machine learning. We give theoretical results describing the algorithm's behavior both on the training data, and on new test data not seen during training. We also describe an efficient implementation of the algorithm for a particular restricted but common case. We next discuss two experiments we carried out to assess the performance of RankBoost. In the first experiment, we used the algorithm to combine different web search strategies, each of which is a query expansion for a given domain. The second experiment is a collaborative-filtering task for making movie recommendations. ML cui01lineage Lineage Tracing for General Data Warehouse Transformations Data warehousing systems integrate information from operational data sources into a central repository to enable analysis and mining of the integrated information. During the integration process, source data typically undergoes a series of transformations, which may vary from simple algebraic operations or aggregations to complex "data cleansing" procedures. In a warehousing environment, the data lineage problem is that of tracing warehouse data items back to the original source items from which they were derived. We formally define the lineage tracing problem in the presence of general data warehouse transformations, and we present algorithms for lineage tracing in this environment. Our tracing procedures take advantage of known structure or properties of transformations when present, but also work in the absence of such information. Our results can be used as the basis for a lineage tracing tool in a general warehousing setting, and also can guide the design of data warehouses that enable efficient lineage tracing. 1 DB 64654 Estimating Dependency Structure as a Hidden Variable This paper introduces a probability model, the mixture of trees that can account for sparse, dynamically changing dependence relationships. We present a family of efficient algorithms that use EM and the Minimum Spanning Tree algorithm to find the ML and MAP mixture of trees for a variety of priors, including the Dirichlet and the MDL priors. 1 INTRODUCTION A fundamental feature of a good model is the ability to uncover and exploit independencies in the data it is presented with. For many commonly used models, such as neural nets and belief networks, the dependency structure encoded in the model is fixed, in the sense that it is not allowed to vary depending on actual values of the variables or with the current case. However, dependency structures that are conditional on values of variables abound in the world around us. Consider for example bitmaps of handwritten digits. They obviously contain many dependencies between pixels; however, the pattern of these dependencies will vary acr... ML eriksson98to To Each and Everyone an Agent: Augmenting Web-Based Commerce with Agents Internet has evolved from an information space to a market space with thousands, potentially millions, of electronic storefronts, auctions and other commercial services. This creates great opportunities, but is not without problems. One major problem is the difficulty of finding relevant offers. Another problem is coping with the multitude of different styles of web-based user interfaces to different marketplaces. Yet another problem is how to automate routine tasks in such an environment. We present one possible solution to these problems. An agent-based market infrastructure, in which agents support all users and services, helps customers and commercial sites find matching interests, and, if desired, negotiate and close deals. The infrastructure is entirely open and decentralized. Each participant has an agent that acts in the interest of its owner. Interaction is entirely symmetric. Any participant can play any role on a market. In this paper we present an integration of such an infrastructure, SICS MarketSpace, with the web. Personal assistant agents help users in their interaction with services and are able to handle routine tasks off-line. Agent-enabled services are able to adapt to the interests of their users, even on their first visit, and are provided with a mechanism to take the first initiative (push) in a highly focused manner. Keywords: agent-based markets, software agents, worldwide web, electronic commerce, personal assistants 1. Agents 455229 Group Task Analysis for Groupware Usability Evaluations Techniques for inspecting the usability of groupware applications have recently been proposed. These techniques focus on the mechanics of collaboration rather than the work context in which a system is used, and offer time and cost savings by not requiring actual users or fully-functional prototypes. Although these techniques are valuable, adding information about task and work context could improve the quality of inspection results. We introduce a method for analysing group tasks that can be used to add context to discount groupware evaluation techniques. Our method allows for the specification of collaborative scenarios and tasks by considering the mechanics of collaboration, levels of coupling during task performance, and variability in task execution. We describe how this type of task analysis could be used in a new inspection technique based on cognitive walkthrough. HCI 291240 An Open Framework for Distributed Multimedia Retrieval This article describes a framework for distributed multimedia retrieval which permits the connection of compliant user interfaces with a variety of multimedia retrieval engines via an open communication protocol, MRML (Multi Media Retrieval Markup Language). It allows the choice of image collection, feature set and query algorithm during run{time, permitting multiple users to query a system adapted to their needs, using the query paradigm adapted to their problem such as query by example (QBE), browsing queries, or query by annotation. User interaction is implemented over several levels and in diverse ways. Relevance feedback is implemented using positive and negative example images that can be used for a best{match QBE query. In contrast, browsing methods try to approach the searched image by giving overviews of the entire collection and by successive renement. In addition to these query methods, Long term o line learning is implemented. It allows feature preferences per ... HCI acharya99selectivity Selectivity Estimation in Spatial Databases Selectivity estimation of queries is an important and wellstudied problem in relational database systems. In this paper, we examine selectivity estimation in the context of Geographic Information Systems, which manage spatial data such as points, lines, poly-lines and polygons. In particular, we focus on point and range queries over two-dimensional rectangular data. We propose several techniques based on using spatial indices, histograms, binary space partitionings (BSPs), and the novel notion of spatial skew. Our techniques carefully partition the input rectangles into subsets and approximate each partition accurately. We present a detailed experimental study comparing the proposed techniques and the best known sampling and parametric techniques. We evaluate them using synthetic as well as real-life TIGER datasets. Based on our experiments, we identify a BSP based partitioning that we call Min-Skew which consistently provides the most accurate selectivity estimates for spatial queries... DB 114091 A Multimodal Approach To Term Extraction Using A Rhetorical Structure Theory Tagger And Formal Concept Analysis This paper reports on knowledge extraction using Rhetorical Structure Theory (RST) and Formal Concept Analysis (FCA). The research is multimodal in two ways: (i) it uses a text tagger to identify key terms in free text, these terms are then used as indexation filters over the free text; (ii) it aims to normalise the contents of multiple text sources into a single knowledge base. The aim is semi-automated extraction of semantic content in texts derived from different sources and merging them into a single coherent knowledge base. We use RST ([7]) to automate the identification of discourse markers in multiple texts dealing with a single subject matter. Marcu ([8, 10]) has shown that RST can be used for the semiautomated mark up of natural language texts. Marcu uses discourse trees, useful to store information about the rhetorical structure, and has shown that the identification of discourse markers from prototypical texts can be automated with 88% precision ([9]). We have adapted Marcu's algorithm in our approach. Although our work draws on recent results from natural language processing, progress in that field is not the objective. The research is motivated by the analysis of texts generated by different sources, their translation to a formal knowledge representation followed by a consolidation into a single knowledge corpus. Our interest is in the analysis of this corpus to determine the reliability of information obtained from multiple agencies ([11]) and then to visually navigate this knowledge. This involves FCA ([14, 15, 17, 18, 6]) for browsing and retrieving text documents ([2, 3, 4, 1]). FCA is typically a propositional knowledge representation technique, i.e., it can only express monadic relations. Recently, Wille ([16]) has shown that FCA can be used to repres... AI palen99social Social, Individual & Technological Issues for Groupware Calendar Systems Designing and deploying groupware is difficult. Groupware evaluation and design are often approached from a single perspective, with a technologically-, individually-, or socially-centered focus. A study of Groupware Calendar Systems (GCSs) highlights the need for a synthesis of these multiple perspectives to fully understand the adoption challenges these systems face. First, GCSs often replace existing calendar artifacts, which can impact users' calendaring habits and in turn influence technology adoption decisions. Second, electronic calendars have the potential to easily share contextualized information publicly over the computer network, creating opportunities for peer judgment about time allocation and raising concerns about privacy regulation. However, this situation may also support coordination by allowing others to make useful inferences about one's schedule. Third, the technology and the social environment are in a reciprocal, co-evolutionary relationship: the use context is affected by the constraints andaffordances of the technology, and the technology also co-adapts to the environment in important ways. Finally, GCSs, despite being below the horizon of everyday notice, can affect the nature of temporal coordination beyond the expected meeting scheduling practice. HCI roth01relational Relational Learning via Propositional Algorithms: An Information Extraction Case Study This paper develops a new paradigm for relational learning which allows for the representation and learning of relational information using propositional means. This paradigm suggests different tradeoffs than those in the traditional approach to this problem -- the ILP approach -- and as a result it enjoys several significant advantages over it. In particular, the new paradigm is more flexible and allows the use of any propositional algorithm, including probabilistic algorithms, within it. We evaluate the new approach on an important and relation-intensive task - Information Extraction - and show that it outperforms existing methods while being orders of magnitude more efficient. 1 IR 441053 Direct value-approximation for factored MDPs We present a simple approach for computing reasonable policies for factored Markov decision processes (MDPs), when the optimal value function can be approximated by a compact linear form. ML prentzas01webbased A Web-Based ITS Controlled by an Expert System Intelligent Tutoring System (ITS) for teaching high school teachers how to use new technologies. It offers course units covering the needs of users with different knowledge levels and characteristics. It tailors the presentation of the educational material to the users' diverse needs by using AI techniques to specify each user's model as well as to make pedagogical decisions. This is achieved via an expert system that uses a hybrid knowledge representation formalism integrating symbolic rules with neurocomputing. AI coetzee00feature Feature Selection in Web Applications Using ROC Inflections and Power Set Pruning A basic problem of information processing is selecting enough features to ensure that events are accurately represented for classification problems, while simultaneously minimizing storage and processing of irrelevant or marginally important features. To address this problem, feature selection procedures perform a search through the feature power set to find the smallest subset meeting performance requirements. Major restrictions of existing procedures are that they typically explicitly or implicitly assume a fixed operating point, and make limited use of the statistical structure of the feature power set. We present a method that combines the Neyman-Pearson design procedure on finite data, with the directed set structure of the Receiver Operating Curves on the feature subsets, to determine the maximal size of the feature subsets that can be ranked in a given problem. The search can then be restricted to the smaller subsets, resulting in significant reductions in computational... IR shehory99emergent Emergent Cooperative Goal-Satisfaction in Large Scale Automated-Agent Systems Cooperation among autonomous agents has been discussed in the DAI community for several years. Papers about cooperation [6, 45], negotiation [33], distributed planning [5], and coalition formation [28, 48], have provided a variety of approaches and several algorithms and solutions to situations wherein cooperation is possible. However, the case of cooperation in large-scale multi-agent systems (MAS) has not been thoroughly examined. Therefore, in this paper we present a framework for cooperative goal-satisfaction in large-scale environments focusing on a low complexity physics-oriented approach. The multi-agent systems with which we deal are modeled by a physics-oriented model. According to the model, MAS inherit physical properties, and therefore the evolution of the computational systems is similar to the evolution of physical systems. To enable implementation of the model, we provide a detailed algorithm to be used by a single agent within the system. The model and the algorithm are a... Agents 5541 Exploiting Schema Knowledge for the Integration of Heterogeneous Sources . Information sharing from multiple heterogeneous sources is a challenging issue which ranges from database to ontology areas. In this paper, we propose an intelligent approach to information integration which takes into account semantic conflicts and contradictions, caused by the lack of a common shared ontology. Our goal is to provide an integrated access to information sources, allowing a user to pose a single query and to receive a single unified answer. We propose a "semantic" approach to integration where the conceptual schema of each source is provided, adopting a common standard data model and language, and Description Logics plus clustering techniques are exploited. Description Logics is used to obtain a semi-automatic generation of a Common Thesaurus (to solve semantic heterogeneities and to derive a common ontology). Clustering techniques are used to build the global schema, i.e. the unified view of the data to be used for query processing. keywords: Intelligent Informat... DB 340027 Improving the Performance of High-Energy Physics Analysis through Bitmap Indices Bitmap indices are popular multi-dimensional structures for accessing read-mostly data such as data warehouse (DW) applications, decision support systems (DSS) and on-line analytical processing (OLAP). One of their main strengths is that they provide good performance characteristics for complex adhoc and an efficient combination of multiple index in one query. Considerable research work has been done in the area of finite (and low) attribute cardinalities. However, additional complexity is imposed on the design of bitmap indices for high cardinality or even non-discrete attributes, where different optimisation techniques than the ones proposed so far have to be applied. In this paper we discuss the design and implementation of bitmap indices for High-Energy Physics (HEP) analysis, where the potential search space consists of hundreds of independent dimensions. A single HEP query typically covers 10 to 100 dimensions out of the whole searchs space. In this context we evaluated two different bitmap encoding techniques, namely equality encoding and range encoding. For both methods the number of bit slices (or bitmap vectors) per attribute is a a central optimisation parameter. The paper presents some (first) results for choosing the optimal number of bit slices for multi-dimensional indices with attributes of different value distribution and query selectivity. We believe taht this discussion is not only applicable to HEP but also to DW, DSS and OLAP type problems in general. DB lisetti00automatic Automatic Facial Expression Interpretation: Where Human-Computer Interaction, Artificial Intelligence and Cognitive Science Intersect this paper is to attempt to bring together people, results and questions from these three different disciplines -- HCI, AI, and Cognitive Science -- to explore the potential of building computer interfaces which understand and respond to the richness of the information conveyed in the human face. Until recently, information has been conveyed from the computer to the user mainly via the visual channel, whereas inputs from the user to the computer have been made from the keyboard and pointing devices via the user's motor channel. The recent emergence of multimodal interfaces as our everyday tools might restore a better balance between our physiology and sensory/motor skills, and impact (for the better we hope), the richness of activities we will find ourselves involved in. Given recent progress in user-interface primitives composed of gesture, speech, context and affect, it seems feasible to design environments which do not impose themselves as computer environments, but have a much more natural feeling associated with them. ML 383150 Logical Structure Detection for Heterogeneous Document Classes We present a fully implemented system based on generic document knowledge for detecting the logical structure of documents for which only general layout information is assumed. In particular, we focus on detecting the reading order. Our system integrates components based on computer vision, articial intelligence, and natural language processing techniques. The prominent feature of our framework is its ability to handle documents from heterogeneous collections. The system has been evaluated on a standard collection of documents to measure the quality of the reading order detection. Experimental results for each component and the system as a whole are presented and discussed in detail. The performance of the system is promising, especially when considering the diversity of the document collection. Keywords: Document Analysis, Logical Structure Detection, Reading Order Detection, Natural Language Processing, Spatial Reasoning. 1. INTRODUCTION The goal of document analysis is to automa... IR dreilinger96information An Information Gathering Agent for Querying Web Search Engines Information gathering agents have attracted much attention of late. As a new application, they are attractive because the need for intelligent assistance in navigating the World Wide Web and large databases is acute. Information agents provide an open-ended and complex, yet easily accessible environment in which ideas from many areas can be integrated. We have developed an information gathering agent called SavvySearch for intelligently searching multiple search engines on the Web. SavvySearch tracks responses from existing search engines to manage resource usage and submit queries only to the most appropriate search engines. To implement SavvySearch, we adapted simple ideas from machine learning, information retrieval and planning and tested two issues in the designs: Can search engine selection knowledge be acquired to improve performance? Do users find that high quality results are being returned early within the limited parallelism provided by SavvySearch? Current results indicate ... IR 346889 Yarrow: A Real-Time Client Side Meta-Search Learner In this paper we report our research on building Yarrow - an intelligent web meta-search engine. The predominant feature of Yarrow is that in contrast to the lack of adaptive learning features in existing metasearch engines, Yarrow is equipped with a practically efficient on-line learning algorithm so that it is capable of helping the user to search for the desired documents with as little feedback as possible. Currently, Yarrow can query eight of the most popular search engines and is able to perform document parsing and indexing, and learning in real-time on client side. Its architecture and performance are also discussed. 1. Introduction As the world wide web evolves and grows so rapidly, web search, an interface between the human users and the vast information gold mine of the web, is becoming a necessary part of people's daily life. Designing and implementing practically effective web search tools is a challenging task. It calls for innovative methods and strategies f... IR baker00hallucinating Hallucinating Faces In most surveillance scenarios there is a large distance between the camera and the objects of interest in the scene. Surveillance cameras are also usually set up with wide fields of view in order to image as much of the scene as possible. The end result is that the objects in the scene normally appear very small in surveillance imagery. It is generally possible to detect and track the objects in the scene, however, for tasks such as automatic face recognition and license plate reading, resolution enhancement techniques are often needed. Although numerous resolution enhancement algorithms have been proposed in the literature, most of them are limited by the fact that they make weak, if any, assumptions about the scene. We propose an algorithm that can be used to learn a prior on the spatial distribution of the image gradient for frontal images of faces. We proceed to show how such a prior can be incorporated into a super-resolution algorithm to yield 4-8 fold improvements in resolution #... ML 538003 Two Views of Classifier Systems This work suggests two ways of looking at Michigan classifier systems; as Genetic Algorithm-based systems, and as Reinforcement Learning-based systems, and argues that the former is more suitable for traditional strength-based systems while the latter is more suitable for accuracy-based XCS. The dissociation of the Genetic Algorithm from policy determination in XCS is noted, and the two types of Michigan classifier system are contrasted with Pittsburgh systems. ML tan01trust Trust Relationships in a Mobile Agent System . The notion of trust is presented as an important component in a security infrastructure for mobile agents. A trust model that can be used in tackling the aspect of protecting mobile agents from hostile platforms is proposed. We dene several trust relationships in our model, and present a trust derivation algorithm that can be used to infer new relationships from existing ones. An example of how such a model can be utilized in a practical system is provided. 1 Agents turner00improving Improving the Scalability of Multi-agent Systems . There is an increasing demand for designers and developers to construct ever larger multi-agent systems. Such systems will be composed of hundreds or even thousands of autonomous agents. Moreover, in open and dynamic environments, the number of agents in the system at any one time will uctuate signicantly. To cope with these twin issues of scalability and variable numbers, we hypothesize that multiagent systems need to be both self-building (able to determine the most appropriate organizational structure for the system by themselves at runtime) and adaptive (able to change this structure as their environment changes). To evaluate this hypothesis we have implemented such a multiagent system and have applied it to the domain of automated trading. Preliminary results supporting the rst part of this hypothesis are presented: adaption and self-organization do indeed make the system better able to cope with large numbers of agents. 1 Introduction When designing or buildin... Agents 460643 Consensus-Based Methods Applied to the Intelligent User Interface Development In today's world properly designed user interfaces are becoming crucial for every information systems. Population of users is very differentiated, so it is almost impossible to design for every information system a single, equally appropriate, user interface for each user. Instead, we postulate to construct adoptive interfaces that take into account experiences of all the population to build users profile by means of consensus methods. Keywords: intelligent user interface, user profile, consensus method 1 HCI billinghurst98wearable A Wearable Spatial Conferencing Space Wearable computers provide constant access to computing and communications resources. In this paper we describe how the computing power of wearables can be used to provide spatialized 3D graphics and audio cues to aid communication. The result is a wearable augmented reality communication space with audio enabled avatars of the remote collaborators surrounding the user. The user can use natural head motions to attend to the remote collaborators, can communicate freely while being aware of other side conversations and can move through the communication space. In this way the conferencing space can support dozens of simultaneous users. Informal user studies suggest that wearable communication spaces may offer several advantages, both through the increase in the amount of information it is possible to access and the naturalness of the interface. 1: Introduction One of the broad trends emerging in human-computer interaction is the increasing portability of computing and communication fac... HCI 499374 A Software Fault Tree Approach to Requirements Analysis of an Intrusion Detection System Requirements analysis for an Intrusion Detection System (IDS) involves deriving requirements for the IDS from analysis of the intrusion domain. When the IDS is, as here, a collection of mobile agents that detect, classify, and correlate system and network activities, the derived requirements include what activities the agent software should monitor, what intrusion characteristics the agents should correlate, where the IDS agents should be placed to feasibly detect the intrusions, and what countermeasures the software should initiate. This paper describes the use of software fault trees for requirements identification and analysis in an IDS. Intrusions are divided into seven stages (following Ruiu), and a fault subtree is developed to model each of the seven stages (reconnaissance, penetration, etc.). Two examples are provided. This approach was found to support requirements evolution (as new intrusions were identified), incremental development of the IDS, and prioritization of countermeasures. Agents biskup00constraints Constraints in Object-Oriented Databases Normal forms in relational database theory, like 3NF or BCNF, are dened by means of semantic contraints. Since for these constraints sound and complete axiomatisations exist and, additionally, for some of these constraints the implication problem is decidable, computer aided database design is possible for relational data models. Object-oriented database theory lacks such normal forms, partly because neither a classication of semantic constraints nor sound and complete axiomatisations exist. In this work we present three classes of semantic constraints for object-oriented data models and show that these constraints have a sound and complete axiomatisation. Thus we prepare the grounds for normal forms in object-oriented data models and subsequently for computer aided object-oriented database design. 1 Introduction The theory of database design for relational data models identies a number of properties to characterise good database schemas. These properties lead then to no... DB 105684 Bargaining with Deadlines This paper analyzes automated distributive negotiation where agents have firm deadlines that are private information. The agents are allowed to make and accept offers in any order in continuous time. We show that the only sequential equilibrium outcome is one where the agents wait until the first deadline, at which point that agent concedes everything to the other. This holds for pure and mixed strategies. So, This material is based upon work supported by the National Science Foundation under CAREER Award IRI-9703122, Grant IRI-9610122, and Grant IIS-9800994. y This is based upon work supported by the EPSRC Award GR M07052. interestingly, rational agents can never agree to a nontrivial split because offers signal enough weakness of bargaining power (early deadline) so that the recipient should never accept. Similarly, the offerer knows that it offered too much if the offer gets accepted: the offerer could have done better by out-waiting the opponent. In most cases, the deadline ... Agents boncz98flattening Flattening an Object Algebra to Provide Performance Algebraic transformation and optimization techniques have been the method of choice in relational query execution, but applying them in OODBMS has been difficult due to the complexity of object-oriented query languages. This paper demonstrates that the problem can be simplified by mapping a complex storage model to the flat binary model implemented by Monet, a state-of-theart database kernel. We present a generic mapping scheme to flatten data models and study the case of a straightforward object-oriented model. We show how flattening enabled us to implement a full-fledged query algebra on it, using only a very limited set of simple operations. The required primitives and query execution strategies are discussed, and their performance is evaluated on the 1GB TPC-D benchmark, showing that our divide-and-conquer approach yields excellent results. 1 Introduction During the last decade, relational database technology has grown towards industrial maturity, and the attention of the research... DB ali98implementing Implementing Schema-theoretic Models of Animal Behavior in Robotic Systems Formal models of animal sensorimotor behavior can provide effective methods for generating robotic intelligence. In this paper we describe how schema-theoretic models of the praying mantis are implemented on a hexapod robot equipped with a real-time color vision system. The model upon which the implementation is based was developed by ethologists studying mantids. This implementation incorporates a wide range of behaviors, including obstacle avoidance, prey acquisition, predator avoidance, mating, and chantlitaxia behaviors. 1 Introduction Ecological robotics refers to incorporating aspects of the relationship a robot maintains with its environment into its control system (i.e., its ecology) [4]. One means for developing such a control system is by exploiting models of behavior developed by ethologists or neuroscientists. Although considerable research has been conducted in the modeling of neural controllers based on animal models (e.g., [3, 5, 14]), incorporation of environmental int... AI 263737 ROL2: A Real Deductive Object-Oriented Database Language . ROL is a strongly typed deductive object-oriented database language. It integrates many important features of deductive databases and object-oriented databases. However, it is only a structurally objectoriented language. In this paper, we describe our extension of ROL called ROL2. ROL2 keeps all the important features of ROL. In addition, it incorporates important behaviorally object-oriented features such as rulebased methods and encapsulation so that it is a real deductive objectoriented database language. It supports object identity, complex objects, class hierarchy, methods, non-monotonic multiple structural and behavioral inheritance with overriding and blocking. 1 Introduction In the past decade, a number of deductive object-oriented database languages have been proposed, such as O-logic [21], revised O-logic [15], IQL [1], LOGRES [7], Datalog meth [2], CORAL++[26], Gulog [10], Rock & Roll [3] Flogic [14], and ROL [18, 19]. However, most of them are only structural... DB voncollani99neurofuzzy A Neuro-Fuzzy Solution for Integrated Visual and Force Control In this paper the use of a B-spline neuro-fuzzy model for different tasks such as vision-based fine-positioning using uncalibrated cameras and force control is presented. It is shown that neuro-fuzzy controllers can be used not only for low-dimensional problems like force control but also for high-dimensional problems like vision-based sensorimotor control and for fusing input from different sensors. Controllers of this type can be modularly combined to solve a given assembly problem. 1 Introduction It is well-known that general fuzzy rule descriptions of systems with a large number of input variables suffer from the problem of the "curse of dimensionality." In many realworld applications it is difficult to identify the decisive input parameters and thus to reduce the number of input variables to the minimum. A general solution to building fuzzy models is not only interesting from a theoretical point, it may also extend the range of applications of fuzzy control to more complex intel... ML liechti99nonobtrusive A Non-obtrusive User Interface for Increasing Social Awareness on the World Wide Web Arguing for the need of increasing social awareness on the World Wide Web, we describe a user interface based on the metaphor of windows bridging electronic and physical spaces. We present a system that, with the aim of making on-line activity perceptible in the physical world, makes it possible to hear people visiting one's Web site. The system takes advantage of the seamless and continuous network connection offered by handheld Web-appliances such as PDA's. IR 261862 The Morph Node We discuss potential and limitations of a Morph Node, inspired by the corresponding construct in Java3D. A Morph Node in Java3D interpolates vertex attributes among several homeomorphic geometries. This node is a promising candidate for the delivery of 3D animation in a very compact form. We review the state-of-the-art in Web 3D techniques, allowing for the possibility of interpolating among several geometries. This review leads to a simple extension for VRML-97 as well as a recommendation for necessary changes in Java3D. Furthermore, we discuss various optimization issues for Morph Nodes. CR Categories and Subject Descriptors. I.3.6 [Computer Graphics ] Methodology and Techniques: Standards - VRML; I.3.7 [Computer Graphics] Three Dimensional Graphics and Realism: Animation; I.3.8 [Computer Graphics] Applications. Additional Keywords. Animation, Avatars, Morphing, Virtual Humans, VRML. INTRODUCTION Animation of three-dimensional shapes involves the change of vertex attributes over ... HCI 518526 HYSSOP: Natural Language Generation Meets Knowledge Discovery in Databases. In this paper, we present HYSSOP, a system that generates natural language hypertext summaries of insights resulting from a knowledge discovery process. We discuss the synergy between the two technologies underlying HYSSOP: Natural Language Generation (NLG) and Knowledge Discovery in Databases (KDD). We first highlight the advantages of natural language hypertext as a summarization medium for KDD results, showing the gains that it provides over charts and tables in terms of conciseness, expressive versatility and ease of interpretation for decision makers. Second, we highlight how KDD technologies, and in particular OLAP and data mining, can implement key tasks of automated natural language data summary generation, in a more domain-independent and scalable way than the human written heuristic rule approach of previous systems. DB bennett00combinations Combinations of Modal Logics Combining logics for modelling purposes has become a rapidly expanding enterprise that is inspired mainly by concerns about modularity and the wish to join together different kinds of information. As any interesting real world system is a complex, composite entity, decomposing its descriptive requirements (for design, verification, or maintenance purposes) into simpler, more restricted, reasoning tasks is not only appealing but is often the only plausible way forward. It would be an exaggeration to claim that we currently have a thorough understanding of `combined methods.' However, a core body of notions, questions and results has emerged for an important class of combined logics, and we are beginning to understand how this core theory behaves when it is applied outside this particular class. In this paper we will consider the combination of modal (including temporal) logics, identifying leading edge research that we, and others, have carried out. Such combined sys Agents gullickson98using Using Experience to Guide Web Server Selection We examine the use of the anycasting communication paradigm to improve client performance when accessing replicated multimedia objects. Anycasting supports dynamic selection of a server amongst a group of servers that provide equivalent content. If the selection is done well, the client will experience improved performance. A key issue in anycasting is the method used to maintain performance information used in server selection. We explore using past performance or experience to predict future performance. We conduct our work in the context of a customized web prefetching application called WebSnatcher. We examine a variety of algorithms for selecting a server using past performance and find that the overall average and weighted average algorithms are closest to optimal performance. In addition to the WebSnatcher application, this work has implications for responsible network behavior by other applications that generate network traffic automatically. By using the techniques we present ... IR 474167 A multi-agent system for advising and monitoring students navigating instructional Web sites A growing community of teachers, at all levels of the educational system, provides course material in the form of hypertext/multimedia documents. In most cases this is done by creating a course Web site. This paper explores the issues related to the design of software systems that aid teachers in monitoring how students use their sites and proactively advise students navigating the sites. In connection to these functions two important topics in current applications of technology to education are discussed. Firstly the definition of a set of criteria allowing the evaluation of the appropriateness of multi-media and hypertext technologies vis vis to classic course support material and in particular textbooks. Secondly the issue of the utility and acceptability of proactive user interfaces such as interface agents or personal assistant agents. A multi agent system capable of advising and monitoring students navigating instructional Web sites is introduced and it is used as a basis for discussion of the above two topics. The system generates and uses a set of indicators evaluating how much use is made of hypertext and multimedia tools as well as indicators of usefulness and cognitive support of the proactive user interface. Keywords: Tutoring systems, Multi agent systems, World Wide Web, Autonomous Interface Agents, Digital Course Material, XML. 1. Agents greiff98theory A Theory of Term Weighting Based on Exploratory Data Analysis Techniques of exploratory data analysis are used to study the weight of evidence that the occurrence of a query term provides in support of the hypothesis that a document is relevant to an information need. In particular, the relationship between the document frequency and the weight of evidence is investigated. A correlation between document frequency normalized by collection size and the mutual information between relevance and term occurrence is uncovered. This correlation is found to be robust across a variety of query sets and document collections. Based on this relationship, a theoretical explanation of the efficacy of inverse document frequency for term weighting is developed which differs in both style and content from theories previously put forth. The theory predicts that a "flattening" of idf at both low and high frequency should result in improved retrieval performance. This altered idf formulation is tested on all TREC query sets. Retrieval results corroborate the predicti... IR lyons00guided Guided by Voices: An Audio Augmented Reality System This paper presents an application of a low cost, lightweight audio-only augmented reality infrastructure. The system uses a simple wearable computer and a RF based location system to play digital sounds corresponding to the user's location and current state. Using this infrastructure we implemented a game in the fantasy genre where players move around in the real world and trigger actions in the virtual game world. We present some of the issues involved in creating audio-only augmented reality games and show how our location infrastructure is generalizable to other audio augmented realities. Keywords Audio, augmented reality, wearable computing, context-awareness INTRODUCTION This paper presents a lightweight and inexpensive infrastructure for augmented realities that uses a simple wearable computer. Whereas most traditional augmented reality systems overlay graphics onto the user's environment, this system employs only audio. Furthermore, we have created a positioning infrastruct... HCI hoste00rule A Rule Induction Approach to Modeling Regional Pronunciation Variation. This 1)~q)er descril)es the use of rule indue-tion techniques fi)r the mli;omatic exl;ra(:l;ion of l)honemic knowledge mM rules fl'om pairs of l:,romm(:intion lexi(:a. This (:xtra(:ted knowl-edge allows the ndat)tntion of sl)ee(:h pro(:ess-ing systelns tO regional vm'iants of a language. As a case sl;u(ty, we apply the approach to Northern Dutch and Flemish (the wtriant of Dutch spoken in Flan(lers, a t)art; of Bel-gium), based Oll C(?lex and l'bnilex, promm-clarion lexi(:a tbr Norttmrn l)utch mM Fhm,-ish, r(~sl)e(:tively. In our study, we (:omt)ar(~ l;wo rule ilMu(:tion techniques, franslbrmation-B;tsed Error-l)riven Learning ('I'I/E])I,) (Brill, 1995) mM C5.0 (Quinl~m, 1993), and (,valu-ate the extr~tct(xl knowh;dge quanl:it~l;ively (a(:-(:ura.cy) mM qualitatively (linguistic r(;levanc:e of the rules). We (:onchMe that. whereas classificntion-1)ased rule. induct;ion with C5.0 is 11101.'0 a(;(:(lr&l;e ~ th(? |;rallSt~)rnl;~l;ion l"ules le;~rne(t with TBE1)I, can 1)e more easily ini;ert)reted. 1. ML dautenhahn99bringing Bringing up Robots or - The Psychology of Socially Intelligent Robots: From Theory to Implementation We discuss robotic experiments in a framework based on theories in developmental psychology. 1 Introduction Piaget's theory of cognitive development has strongly influenced many approaches in artificial intelligence and agent research. His theory has been challenged from various directions, and recent experiments have confirmed Vygotsky's belief in the essential role of social interaction and teaching as a scaffolding mechanism which is important for the child in order to reach higher levels of competence and control based on current skills. Hereby concepts are not taught directly but through social interaction, the child's experiences are re-arranged, a shared understanding develops between the child and its interaction partner. A Piagetean viewpoint sees language as a product of the cognitive development of mental representations, while Vygotsky believes that the sole primary function of language is communication with peers and adults, and that language develops exactly in this cont... AI argamon98routing Routing Documents According to Style Most research on automated text categorization has focused on determining the topic of a given text. While topic is generally the main characteristic of an information need, there are other characteristics that are useful for information retrieval. In this paper we consider the problem of text categorization according to style. For example, in searching the web, we may wish to automatically determine if a given page is promotional or informative, was written by a native English speaker or not, and so on. Learning to determine the style of a document is a dual to that of determining its topic, in that those document features which capture the style of a document are precisely those which are independent of its topic. We here define the features of a document to be the frequencies of each of a set of function words and parts-of-speech triples. We then use machine learning techniques to classify documents. We test our methods on four collections of downloaded newspaper and magazine articl... IR 35592 Boosting the Margin: A New Explanation for the Effectiveness of Voting Methods Abstract. One of the surprising recurring phenomena observed in experiments with boosting is that the test error of the generated hypothesis usually does not increase as its size becomes very large, and often is observed to decrease even after the training error reaches zero. In this paper, we show that this phenomenon is related to the distribution of margins of the training examples with respect to the generated voting classification rule, where the margin of an example is simply the difference between the number of correct votes and the maximum number of votes received by any incorrect label. We show that techniques used in the analysis of Vapnik’s support vector classifiers and of neural networks with small weights can be applied to voting methods to relate the margin distribution to the test error. We also show theoretically and experimentally that boosting is especially effective at increasing the margins of the training examples. Finally, we compare our explanation to those based on the bias-variance decomposition. 1 ML ordonez01mining Mining Constrained Association Rules to Predict Heart Disease This work describes our experiences on discovering association rules in medical data to predict heart disease. We focus on two aspects in this work: mapping medical data to a transaction format suitable for mining association rules and identifying useful constraints. Based on these aspects we introduce an improved algorithm to discover constrained association rules. We present an experimental section explaining several interesting discovered rules. 1. AI svidzinska01world A World Wide Web Meta Search Engine Using an Automatic Query Routing Algorithm CONTENTS 1. INTRODUCTION......................................................... ........................................ 6 2. LITERATURE REVIEW............................................................... ........................ 9 2.1 Overview of conventional search techniques............................................. 9 2.2 Conventional query routing systems......................................................... 11 2.2.1 Manual query routing services......................................................... 11 2.2.2 Automated query routing systems based on centroids..................... 12 2.2.3 Automated query routing systems without centroids....................... 12 3. SYSTEM STRUCTURE............................................................ .......................... 14 3.1 System overview............................................................. .......................... 14 3.2 Off-line operations....................................................... IR duch98optimization Optimization and Global Minimization Methods Suitable for Neural Networks Neural networks are usually trained using local, gradient-based procedures. Such methods frequently find suboptimal solutions being trapped in local minima. Optimization of neural structures and global minimization methods applied to network cost functions have strong influence on all aspects of network performance. Recently genetic algorithms are frequently combined with neural methods to select best architectures and avoid drawbacks of local minimization methods. Many other global minimization methods are suitable for that purpose, although they are used rather rarely in this context. This paper provides a survey of such global methods, including some aspects of genetic algorithms. CONTENTS 1 Introduction 2 2 Monte Carlo and its improvements 4 3 Simulated annealing and its variants 6 3.1 Adaptive Simulated Annealing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 3.2 Alopex . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... ML tourapis01advanced Advanced DeInterlacing techniques with the use of Zonal Based Algorithms This paper describes a new highly efficient aleinterlacing approach based on motion estimation and compensation techniques. The proposed technique mainly benefits from the motion vector properties of zonal based algorithms, such as the Advanced Predictive Diamond Zonal Search (APDZS) and the Predictive Motion Vector Field Adaptive Search Technique (PMVFAST), multihypothesis motion compensation, but also an additional motion classification phase where, depending on the motion of a pixel, additional spatial and temporal information is also considered to further improve performance. Extensive simulations demonstrate the efficacy of these algorithms, especially when compared to standard deinterlacing techniques such as the line doubling and line averaging algorithms. ML 488749 LivingLab: A white paper The LivingLab is a planned research infrastructure that is pivotal for user-system interaction research in the next decade. This article presents the concept and outlines a research programme that will be served by this facility. These future plans are motivated by a vision of future developments concerning interaction with intelligent environments. HCI tenhagen01continuous Continuous State Space Q-Learning for Control of Nonlinear Systems Contents 1 Introduction 1 1.1 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1.1 Designing the state feedback controller . . . . . . . . . . . . . . . . 3 1.1.2 Unknown systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.2 Reinforcement Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.3 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.4 Overview of this Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 2 Reinforcement Learning 11 2.1 A Discrete Deterministic Optimal Control Task . . . . . . . . . . . . . . . 11 2.1.1 The problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2.1.2 The solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.2 The Stochastic Optimization Tasks . . . . . . . . . . . . . . . . . . . . . . 13 2.2.1 The Markov Decision Process . . . . . . . . . . . . . . . . . . . ML deloach02analysis Analysis and Design of Multiagent Systems Using Hybrid Coordination Media Over the last few years, two advances in agent-oriented software engineering have had a significant impact. The first is the identification of interaction and coordination as the central focus of multiagent systems design and the second is the realization that the multiagent organization is distinct from the individual agents that populate the system. Also, the evolution of new, more powerful hybrid coordination models, which combine data-centered and control-centered coordination approaches, have given us the capability to model and implement the rules that govern organizations independently from the individual agents in the system. This paper investigates how to combine the power of these hybrid coordination capabilities with the concept of organizational rules using traditional conversation-based approaches to designing multiagent systems. 1. Agents hespanha99multipleagent Multiple-Agent Probabilistic Pursuit-Evasion Games In this paper we develop a probabilistic framework for pursuit-evasion games. We propose a "greedy" policy to control a swarm of autonomous agents in the pursuit of one or several evaders. At each instant of time this policy directs the pursuers to the locations that maximize the probability of finding an evader at that particular time instant. It is shown that, under mild assumptions, this policy guarantees that an evader is found in finite time and that the expected time needed to find the evader is also finite. Simulations are included to illustrate the results. 1 Introduction This paper addresses the problem of controlling a swarm of autonomous agents in the pursuit of one or several evaders. To this effect we develop a probabilistic framework for pursuit-evasion games involving multiple agents. The problem is nondeterministic because the motions of the pursuers/evaders and the devices they use to sense their surroundings require probabilistic models. It is also assumed that when ... Agents 367649 Let's Browse: A Collaborative Browsing Agent Web browsing, like most of today's desktop applications, is usually a solitary activity. Other forms of media, such as watching television, are often done by groups of people, such as families or friends. What would it be like to do collaborative Web browsing? Could the computer provide assistance to group browsing by trying to help find mutual interests among the participants? Let's Browse is an experiment in building an agent to assist a group of people in browsing, by suggesting new material likely to be of common interest. It is built as an extension to the single user Web browsing agent Letizia. Let's Browse features automatic detection of the presence of users, automated "channel surfing" browsing, and dynamic display of the user profiles and explanation of recommendations. # 1999 Elsevier Science B.V. All rights reserved. Keywords: Browsing; Collaboration; Agents; User profiles 1. Collaborative browsing Increasingly, Web browsing will be performed in collaborative settings, ... HCI milo00typechecking Typechecking for XML Transformers We study the typechecking problem for XML transformers: given an XML transformation program and a DTD for the input XML documents, check whether every result of the program conforms to a specified output DTD. We model XML transformers using a novel device called a k- pebble transducer, that can express most queries without data-value joins in XML-QL, XSLT, and other XML query languages. Types are modeled by regular tree languages, a robust extension of DTDs. The main result of the paper is that typechecking for k-pebble transducers is decidable. Consequently, typechecking can be performed for a broad range of XML transformation languages, including XMLQL and a fragment of XSLT. 1. INTRODUCTION Traditionally, database query languages have focused on data retrieval, with complex data transformations left to applications. The new XML data exchange standard for the Web, and emerging applications requiring data wrapping and integration, have shifted the focus towards data transformations.... DB huntsberger98bismarc BISMARC: A Biologically Inspired System for Map-based Autonomous Rover Control As the complexity of the missions to planetary surfaces increases, so too does the need for autonomous rover systems. This need is complicated by the power, mass and computer storage restrictions on such systems (Miller, 1992). To address these problems, we have recently developed a system called BISMARC (Biologically Inspired System for Map-based Autonomous Rover Control) for planetary missions involving multiple small, lightweight surface rovers (Huntsberger, 1997). BISMARC is capable of cooperative planetary surface retrieval operations such as a multiple cache recovery mission to Mars. The system employs autonomous navigation techniques, behavior-based control for surface retrieval operations, and an action selection mechanism based on a modified form of free flow hierarchy (Rosenblatt and Payton, 1989). This paper primarily describes the navigation and map-mapping subsystems of BISMARC. They are inspired by some recent studies of London taxi drivers indicating that the right hippo... AI lenzmann97contractnetbased Contract-Net-Based Learning in a User-Adaptive Interface Agency . This paper describes a multi-agent learning approach to adaptation to users' preferences realized by an interface agency. Using a contract-net-based negotiation technique, agents as contractors as well as managers negotiate with each other to pursue the overall goal of dynamic user adaptation. By learning from indirect user feedback, the adjustment of internal credit vectors and the assignment of contractors that gained maximal credit with respect to the user's current preferences, the preceding session, and current situational circumstances can be realized. In this way, user adaptation is achieved without accumulating explicit user models but by the use of implicit, distributed user models. 1 Introduction Interface agents are computer programs that enhance the human-computer interaction by mediating a relationship between technical systems and users [Lau90]. On the one hand, they provide assistance to users by acting on his/her behalf and automating his/her actions [Nor94... HCI leuschel98creating Creating Specialised Integrity Checks Through Partial Evaluation Of Meta-Interpreters Interpretation". Danny De Schreye is senior research associate of the Belgian National Fund for Scientific Research. We would like to thank Bern Martens for proof-reading several versions of this paper and for his helpful insights and comments on the topic of this paper. We would also like to thank him for his huge pile of references on integrity checking and for introducing the first author to the subject. We thank Bart Demoen for sharing his expertise on writing efficient Prolog programs. Our thanks also go to John Gallagher for pointing out several errors in an earlier version of the paper and for the fruitful discussions on partial evaluation and integrity checking. We are also grateful for discussions and extremely helpful comments by Hendrik Decker. Other interesting discussions about the topic of this paper were held with Stefan Decker, the members of the Compulog II project as well as with the participants of the 1996 Dagstuhl seminar on "Logic and the Meaning of Change". Fina... DB 131669 On the Correspondence between Neural Folding Architectures and Tree Automata The folding architecture together with adequate supervised training algorithms is a special recurrent neural network model designed to solve inductive inference tasks on structured domains. Recently, the generic architecture has been proven as a universal approximator of mappings from rooted labeled ordered trees to real vector spaces. In this article we explore formal correspondences to the automata (language) theory in order to characterize the computational power (representational capabilities) of different instances of the generic folding architecture. As the main result we prove that simple instances of the folding architecture have the computational power of at least the class of deterministic bottom-up tree automata. It is shown how architectural constraints like the number of layers, the type of the activation functions (first-order vs. higher-order) and the transfer functions (threshold vs. sigmoid) influence the representational capabilities. All proofs are carried out in a c... ML matzinger98computational On Computational Representations of Herbrand Models . Finding computationally valuable representations of models of predicate logic formulas is an important issue in the field of automated theorem proving, e.g. for automated model building or semantic resolution. In this article we treat the problem of representing single models independently of building them and discuss the power of different mechanisms for this purpose. We start with investigating context-free languages for representing single Herbrand models. We show their computational feasibility and prove their expressive power to be exactly the finite models. We show an equivalence with "ground atoms and ground equations" concluding equal expressive power. Finally we indicate how various other well known techniques could be used for representing essentially infinite models (i.e. models of not finitely controllable formulas), thus motivating our interest in relating model properties with syntactical properties of corresponding Herbrand models and in investigating connections betwe... DB baltsavias00integrating Integrating Spatial Information And Image Analysis - One Plus One Makes Ten Photogrammetry and remote sensing have proven their efficiency for spatial data collection in many ways. Interactive mapping at digital workstations is performed by skilled operators, which guarantees excellent quality in particular of the geometric data. In this way, worldwide acquisition of a large number of national GIS databases has been supported and still a lot of production effort is devoted to this task. In the field of image analysis, it has become evident that algorithms for scene interpretation and 3D reconstruction of topographic objects, which rely on a single data source, cannot function efficiently. Research in two directions promises to be more successful. Multiple, largely complementary, sensor data like range data from laser scanners, SAR and panchromatic or multi-/hyper-spectral aerial images have been used to achieve robustness and better performance in image analysis. On the other hand, given GIS databases, e.g. layers from topographic maps, can be considered as vi... DB 483730 QUEST - Querying specialized collections on the Web Ensuring access to specialized web-collections in a fast evolving web environment requires flexible techniques for orientation and querying. The adoption of meta search techniques for web-collections is hindered by the enormous heterogeneity of the resources. In this paper we introduce QUEST --- a system for querying specialized collections on the web. IR frank99naive Naive Bayes for Regression Abstract. Despite its simplicity, the naive Bayes learning scheme performs well on most classification tasks, and is often significantly more accurate than more sophisticated methods. Although the probability estimates that it produces can be inaccurate, it often assigns maximum probability to the correct class. This suggests that its good performance might be restricted to situations where the output is categorical. It is therefore interesting to see how it performs in domains where the predicted value is numeric, because in this case, predictions are more sensitive to inaccurate probability estimates. This paper shows how to apply the naive Bayes methodology to numeric prediction (i.e., regression) tasks by modeling the probability distribution of the target value with kernel density estimators, and compares it to linear regression, locally weighted linear regression, and a method that produces “model trees”—decision trees with linear regression functions at the leaves. Although we exhibit an artificial dataset for which naive Bayes is the method of choice, on real-world datasets it is almost uniformly worse than locally weighted linear regression and model trees. The comparison with linear regression depends on the error measure: for one measure naive Bayes performs similarly, while for another it is worse. We also show that standard naive Bayes applied to regression problems by discretizing the target value performs similarly badly. We then present empirical evidence that isolates naive Bayes ’ independence assumption as the culprit for its poor performance in the regression setting. These results indicate that the simplistic statistical assumption that naive Bayes makes is indeed more restrictive for regression than for classification. ML ashri00paradigma Paradigma: Agent Implementation through Jini One of the key problems of recent years has been the divide between theoretical work in agent-based systems and its practical complement which have, to a large extent, developed along different paths. The Paradigma implementation framework has been designed with the aim of narrowing this gap. It relies on an extensive formal agent framework implemented using recent advances in Java technology. Specifically, Paradigma uses Jini connectivity technology to enable the creation of on-line communities in support of the development of agent-based systems. 1 Introduction In a networked environment that is highly interconnected, interdependent and heterogeneous, we are faced with an explosion of information and available services that are increasingly hard to manage. Agent-based systems can provide solutions to these problems as a consequence of their dynamics of social interaction; communication and cooperation can be used to effectively model problem domains through the interaction of agent... AI goasdoue00rewriting Rewriting Conjunctive Queries Using Views in Description Logics with Existential Restrictions this paper, extending the work of [8], we study the problem of rewriting conjunctive queries over DL expressions into conjunctive queries using a set of views that are a set of distinguished DL expressions, for three DLs allowing existential restrictions: FLE , ALE and ALNE . Thus, our rewriting problem is: given a conjunctive query over expressions from a DL L 2 fFLE;ALE ; ALNEg and a set of views V over expressions from L, we want to compute a representative set of all the rewritings of the query that are conjunctive queries over V. By representative set we mean that this set contains at least the rewritings that are maximally contained in the query. DB davison98applying Applying Parallelism to Improve Genetic Algorithm-based Design Optimization Introduction The abundance of powerful workstations makes course-grained parallelization an obvious enhancement to many optimization techniques, including genetic algorithms [Gol89, DM97]. While initial modifications have been made to GADO (Genetic Algorithm for Design Optimization [Ras98, RHG97]), such changes have not been carefully analyzed for potential impacts on quality. More generally, parallelization has the potential to improve GA performance through the use of alternative models of computation. Parallelism can certainly reduce the total elapsed clock-time for a solution, but as a change in model of computation (either real or simulated) , it can change the number of simulator calls and even make new solutions achievable. The effects of parallelization on GADO were investigated during my summer internship at the Center for Computational Design. 2 Objectives Since a straightforward parallelized implementation already existed, my first tasks were to ana ML 79013 Using the CONDENSATION Algorithm for Robust, Vision-based Mobile Robot Localization To navigate reliably in indoor environments, a mobile robot must know where it is. This includes both the ability of globally localizing the robot from scratch, as well as tracking the robot's position once its location is known. Vision has long been advertised as providing a solution to these problems, but we still lack efficient solutions in unmodified environments. Many existing approaches require modification of the environment to function properly, and those that work within unmodified environments seldomly address the problem of global localization. In this paper we present a novel, vision-based localization method based on the CONDENSATION algorithm [17, 18], a Bayesian filtering method that uses a samplingbased density representation. We show how the CONDEN- SATION algorithm can be used in a novel way to track the position of the camera platform rather than tracking an object in the scene. In addition, it can also be used to globally localize the camera platform, given a visua... ML panayiotopoulos98intelligent An Intelligent Agent Framework In VRML Worlds actions, e.g. move to next room, are received by it and consequently send to the EAC. Finally, the abstract action arrives at the Virtual Reality Management Unit that specifies in detail the received actions. It provides specific values concerning the orientation and position of the avatar, e.g. it specifies the coordinates, orientation and path so that it can successfully move to the next room, and sends them as commands to the Virtual Reality World Browser. The browser executes the command by altering the virtual environment appropriately. When changes have been performed the AEC unit notifies the logical core that the action has been successfully executed and the logical core goes on by updating its internal and external state. Consequently, the agent looks around into the virtual space, gathers any additional information and decides the next step it should take to satisfy its goals. Agents 467872 Model-Free Least-Squares Policy Iteration We propose a new approach to reinforcement learning which combines least squares function approximation with policy iteration. Our method is model-free and completely off policy. We are motivated by the least squares temporal difference learning algorithm (LSTD), which is known for its efficient use of sample experiences compared to pure temporal difference algorithms. LSTD is ideal for prediction problems, however it heretofore has not had a straightforward application to control problems. Moreover, approximations learned by LSTD are strongly influenced by the visitation distribution over states. Our new algorithm, Least-Squares Policy Iteration (LSPI) addresses these issues. The result is an off-policy method which can use (or reuse) data collected from any source. We test LSPI on several problems, including a bicycle simulator in which it learns to guide the bicycle to a goal efficiently by merely observing a relatively small number of completely random trials. ML gurret99basis The BASIS System: a Benchmarking Approach for Spatial Index Structures This paper describes the design of the BASIS prototype system, which is currently under implementation. BASIS stands for Benchmarking Approach for Spatial Index Structures. It is a prototype system aiming at performance evaluation of spatial access methods and query processing strategies, under different data sets, various query types, and different workloads. BASIS is based on a modular architecture, composed of a simple storage manager, a query processor, and a set of algorithmic techniques to facilitate benchmarking. The main objective of BASIS is twofold: (i) to provide a benchmarking environment for spatial access methods and related query evaluation techniques, and (ii) to allow comparative studies of spatial access methods in different cases but under a common framework. We currently extend it to support the fundamental features of spatiotemporal data management and access methods. DB antifakos01exploration Exploration of Perceptual Computing for Smart-Its The future success of ubiquitous computing depends to a big part on how well applications can adapt to their environment and act accordingly. This thesis has set itself the goal of exploring perceptual computing for Smart-Its, which is one such ubiquitous computing vision. HCI barbosa01tox ToX - The Toronto XML Engine Abstract. We present ToX – the Toronto XML Engine – a repository for XML data and metadata, which supports real and virtual XML documents. Real documents are stored as files or mapped into relational or object databases, depending on their structuredness; indices are defined according to the storage method used. Virtual documents can be remote documents, defined as arbitrary WebOQL queries, or views, defined as queries over documents registered in the system. The system catalog contains metadata for the documents, especially their schemata, used for query processing and optimization. Queries can range over both the catalog and the documents, and multiple query languages are supported. In this paper we describe the architecture and main of ToX; we present our indexing and storage strategies, including two novel techniques; and we discuss our query processing strategy. The project started recently and is under active development. 1 DB 68623 Logic-Based Subsumption Architecture In this paper we describe a logic-based AI architecture based on Brooks' Subsumption Architecture. We axiomatize each of the layers of control in his system separately and use independent theorem provers to derive each layer's output actions given its inputs. We implement the subsumption of lower layers by higher layers using circumscription. We give formal semantics to our approach. 1 Introduction In [?], Brooks proposed a reactive architecture embodying an approach to robot control different on various counts from traditional approaches. He decomposed the problem into layers corresponding to levels of behavior, rather than according to a sequential, functional form. Within this setting he introduced the idea of subsumption, that is, that more complex layers could not only depend on lower, more reactive layers, but could also influence their behavior. The resulting architecture was one that could service simultaneously multiple, potentially conflicting goals in a reactive fashi... AI 531929 REFEREE: An open framework for practical testing of recommender systems using ResearchIndex Automated recommendation (e.g., personalized product recommendation on an ecommerce web site) is an increasingly valuable service associated with many databases--typically online retail catalogs and web logs. Currently, a major obstacle for evaluating recommendation algorithms is the lack of any standard, public, real-world testbed appropriate for the task. In an attempt to fill this gap, we have created REFEREE, a framework for building recommender systems using ResearchIndex--a huge online digital library of computer science research papers--so that anyone in the research community can develop, deploy, and evaluate recommender systems relatively easily and quickly. Research Index is in many ways ideal for evaluating recommender systems, especially so-called hybrid recommenders that combine information filtering and collaborative filtering techniques. The documents in the database are associated with a wealth of content information (author, title, abstract, full text) and collaborative information (user behaviors), as well as linkage information via the citation structure. Our framework supports more realistic evaluation metrics that assess user buy-in directly, rather than resorting to offline metrics like prediction accuracy that may have little to do with end user utility. The sheer scale of ResearchIndex (over 500,000 documents with thousands of user accesses per hour) will force algorithm designers to make real-world trade-offs that consider performance, not just accuracy. We present our own tradeoff decisions in building an example hybrid recommender called PD-Live. The algorithm uses content-based similarity information to select a set of documents from which to recommend, and collaborative information to rank the documents. PD-Live performs reasonably well compared to other recommenders in ResearchIndex. IR 386535 Performance Analysis of Mobile Agents for Filtering Data Streams on Wireless Networks Wireless networks are an ideal environment for mobile agents, since their mobility allows them to move across an unreliable link to reside on a wired host, next to or closer to the resources that they need to use. Furthermore, clientspecific data transformations can be moved across the wireless link and run on a wired gateway server, reducing bandwidth demands. In this paper we examine the tradeoffs faced when deciding whether to use mobile agents in a datafiltering application where numerous wireless clients filter information from a large data stream arriving across the wired network. We develop an analytical model and use parameters from filtering experiments conducted during a U.S. Navy Fleet Battle Experiment (FBE) to explore the model's implications. 1. Introduction Mobile agents are programs that can migrate from host to host in a network of computers, at times and to places of their own choosing. Unlike applets, both the code and the execution state (heap and stack) move with... Agents 157253 Latent Semantic Indexing: A Probabilistic Analysis Latent semantic indexing (LSI) is an information retrieval technique based on the spectral analysis of the term-document matrix, whose empirical success had heretofore been without rigorous prediction and explanation. We prove that, under certain conditions, LSI does succeed in capturing the underlying semantics of the corpus and achieves improved retrieval performance. We also propose the technique of random projection as a way of speeding up LSI. We complement our theorems with encouraging experimental results. We also argue that our results may be viewed in a more general framework, as a theoretical basis for the use of spectral methods in a wider class of applications such as collaborative filtering. Computer Science Division, U. C. Berkeley, Berkeley, CA 94720. Email: christos@cs.berkeley.edu. y IBM Almaden Research Center, 650 Harry Road, San Jose, CA 95120. Email: pragh@almaden.ibm.com. z Computer Science Department, Meiji University, Tokyo, Japan. Email: htamaki@cs.meiji.... DB dahl99eel The Eel Programming Language and Internal Concurrency in Logic Agents This paper describes work done on creating the logic programming language Eel. The language is designed for implementing agents with a behaviour based, concurrent internal architecture. The paper also suggests a new such architecture which improves on the ones currently available. It gives examples of how parts of that architecture are implemented in Eel and comments that the Agent Oriented Programming paradigm currently contains two different metaphors for concurrency. Eel's event based approach to process communication and process initiation introduces an explicit representation of state to a logic program. A new declarative approach to object states is demonstrated as a part of the object oriented implementation of the suggested agent architecture. As well as being a programming language, Eel is a formalism which is well suited for logic based machine learning of behaviour and interaction. This paper briefly outlines the scope for using such learning to improve on exist... Agents rui99image Image Retrieval: Current Techniques, Promising Directions And Open Issues This paper provides a comprehensive survey of the technical achievements in the research area of image retrieval, especially content-based image retrieval, an area that has been so active and prosperous in the past few years. The survey includes 100+ papers covering the research aspects of image feature representation and extraction, multidimensional indexing, and system design, three of the fundamental bases of content-based image retrieval. Furthermore, based on the state-of-the-art technology available now and the demand from real-world applications, open research issues are identified and future promising research directions are suggested. C ○ 1999 Academic Press 1. IR 157013 Multi-Layer Incremental Induction . This paper describes a multi-layer incremental induction algorithm, MLII, which is linked to an existing nonincremental induction algorithm to learn incrementally from noisy data. MLII makes use of three operations: data partitioning, generalization and reduction. Generalization can either learn a set of rules from a (sub)set of examples, or refine a previous set of rules. The latter is achieved through a redescription operation called reduction: from a set of examples and a set of rules, we derive a new set of examples describing the behaviour of the rule set. New rules are extracted from these behavioral examples, and these rules can be seen as meta-rules, as they control previous rules in order to improve their predictive accuracy. Experimental results show that MLII achieves significant improvement on the existing nonincremental algorithm HCV used for experiments in this paper, in terms of rule accuracy. 1 Introduction Existing machine learning algorithms can be generally distin... ML 50068 Super Logic Programs Recently, considerable interest and research e#ort has been given to the problem of finding a suitable extension of the logic programming paradigm beyond the class of normal logic programs. In order to demonstrate that a class of programs can be justifiably called an extension of logic programs one should be able to argue that: . the proposed syntax of such programs resembles the syntax of logic programs but it applies to a significantly broader class of programs; . the proposed semantics of such programs constitutes an intuitively natural extension of the semantics of normal logic programs; . there exists a reasonably simple procedural mechanism allowing, at least in principle, to compute the semantics; . the proposed class of programs and their semantics is a special case of a more general non-monotonic formalism which clearly links it to other well-established non-monotonic formalisms. In this paper we propose a specific class of extended logic programs which will be (modestly) called super logic programs or just super-programs. We will argue that the class of super-programs satisfies all of the above conditions, and, in addition, is su#ciently flexible to allow various application-dependent extensions and modifications. We also provide a brief description of a Prolog implementation of a query-answering interpreter for the class of super-programs which is available via FTP and WWW. Keywords: Non-Monotonic Reasoning, Logics of Knowledge and Beliefs, Semantics of Logic Programs and Deductive Databases. # An extended abstract of this paper appeared in the Proceedings of the Fifth International Conference on Principles of Knowledge Representation and Reasoning (KR'96), Boston, Massachusetts, 1996, pp. 529--541. + Partially supported by the National Science Fou... AI 73962 Supporting Flexibility. A Case-Based Reasoning Approach This paper presents a case-based reasoning system TA3. We address the flexibility of the case-based reasoning process, namely flexible retrieval of relevant experiences, by using a novel similarity assessment theory. To exemplify the advantages of such an approach, we have experimentally evaluated the system and compared its performance to the performance of non-flexible version of TA3 and to other machine learning algorithms on several domains. Introduction Flexible Computation There are many situation when resources are scarce and when the system has to make decision as to how to proceed with further computation in order to make the right tradeoff between quality of the answer and resources needed. In several domains, such as medicine, robotics, financing, etc., if the answer to a query is not produced within a certain time limit it may become useless. Such family of problems has been a motivation for designing anytime algorithms (Dean & Boddy 1988; Frisch & Haddawy 1994), algorit... ML 254693 Indexing Moving Points We propose three indexing schemes for storing a set S of N points in the plane, each moving along a linear trajectory, so that a query of the following form can be answered quickly: Given a rectangle R and a real value t q , report all K points of S that lie inside R at time t q . We first present an indexing structure that, for any given constant " ? 0, uses O(N=B) disk blocks, where B is the block size, and answers a query in O((N=B) 1=2+" + K=B) I/Os. It can also report all the points of S that lie inside R during a given time interval. A point can be inserted or deleted, or the trajectory of a point can be changed, in O(log 2 B N) I/Os. Next, we present a general approach that improves the query time if the queries arrive in chronological order, by allowing the index to evolve over time. We obtain a tradeoff between the query time and the number of times the index needs to be updated as the points move. We also describe an indexing scheme in which the number of I/Os required to answer a query depends monotonically on the difference between t q and the current time. Finally, we develop an efficient indexing scheme to answer approximate nearest-neighbor queries among moving points. An extended abstract of this paper appeared in the Proceedings of the 19th ACM SIGACT-SIGMOD-SIGART Symposium on Principles of Database Systems. y Center for Geometric Computing, Department of Computer Science, Duke University, Box 90129, Durham, NC 27708--0129; pankaj@cs.duke.edu; http://www.cs.duke.edu/ pankaj. Supported in part by National Science Foundation grants EIA--9870734, EIA--9972879, and CCR--9732787, by Army Research Of fice MURI grant DAAH04-- 96--1--0013, by a Sloan fellowship, and by a grant from the U.S.-Israeli Binational Science Foundation. z Center ... DB dejong01multilayer Multi-Layer Methods and the Optimal Optimizer Multi-Layer Methods are methods that act on several layers simultaneously. Examples of multi-layer methods are found in multi-agent systems (global and per-agent behavior), in learning (e.g. boosting, bias tuning), in self-adaptive methods (such as evolution strategies), in hybrid approaches, and in optimization (e.g. multiple runs, result pooling). We give a formal definition of what a multi-layer method is. We discuss the relationship with the no free lunch theorem, to show that such a thing as the optimal optimizer exists, and how multi-layer methods can be used to approximate it. Agents hanson99scalable Scalable Trigger Processing + Current database trigger systems have extremely limited scalability. This paper proposes a way to develop a truly scalable trigger system. Scalability to large numbers of triggers is achieved with a trigger cache to use main memory effectively, and a memory-conserving selection predicate index based on the use of unique expression formats called expression signatures. A key observation is that if a very large number of triggers are created, many will have the same structure, except for the appearance of different constant values. When a trigger is created, tuples are added to special relations created for expression signatures to hold the trigger's constants. These tables can be augmented with a database index or main-memory index structure to serve as a predicate index. The design presented also uses a number of types of concurrency to achieve scalability, including token (tuple)-level, condition-level, rule action-level, and datalevel concurrency. 1. Introduction Trigger feature... DB 332050 Agent-Based Modeling for Holonic Manufacturing Systems with Fuzzy Control Agent-based systems technologies are of emerging interest in the specification and implementation of complex systems. This article introduces the CASA agent development system which seamlessly combines the BDI (Belief Desire Intention) approach with the FIPA agent communication language standard and an integrated specification of fuzzy controllers. The behavior of agents is defined by strategies which basically correspond to extended guarded horn clauses with priorities. The presented concepts are introduced by an example from Computer Integrated Manufacturing (CIM). The example gives the specification of a fuzzy controller for a manufacturing station in the context of a holonic manufacturing system (HMS). 1 Introduction Agent-based systems technologies in the sense of distributed computing is an area of emerging interest in the domain of complex systems design. The agent-based paradigm can be seen as a real enhancement of the objectoriented paradigm where objects become autonomous, ... Agents hero98robust Robust Entropy Estimation Strategies Based on Edge Weighted Random Graphs (with corrections) In this paper we treat the problem of robust entropy estimation given a multidimensional random sample from an unknown distribution. In particular, we consider estimation of the Renyi entropy of fractional order which is insensitive to outliers, e.g. high variance contaminating distributions, using the k-point minimal spanning tree (kMST) . A greedy algorithm for approximating the NP-hard problem of computing the k-minimal spanning tree is given which is a generalization of the potential function partitioning method of Ravi etal. 1 The basis for our approach is an asymptotic theorem establishing that the log of the overall length or weight of the greedy approximation is a strongly consistent estimator of the Renyi entropy. Quantitative robustness of the estimator to outliers is established using Hampel's method of inuence functions. 2 The structure of the inuence function indicates that the k-MST is a natural extension of the one dimensional -trimmed mean for multi-dimensional... ML bouguettaya99using Using Java and CORBA for Implementing Internet Databases We describe an architecture called WebFINDIT that allows dynamic couplings of Web accessible databases based on their content and interest. We propose an implementation using WWW, Java, JDBC, and CORBA's ORBs that communicate via the CORBA's IIOP protocol. The combination of these technologies offers a compelling middleware infrastructure to implement wide-area enterprise applications. In addition to a discussion of WebFINDIT's core concepts and implementation architecture, we also discuss an experience of using WebFINDIT in a healthcare application. 1 Introduction The growth of the Internet and the Web increased dramatically the need for data sharing. The Web has brought a wave of new users and service providers to the Internet. It contains a huge quantity of heterogeneous information and services (e.g., home pages, online digital libraries, product catalogs, and so on) (Bouguettaya et al. 1998). The result is that the Web is now accepted as the de facto support in all domains of li... DB 487862 Cross Entropy Guided Ant-like Agents Finding Dependable Primary/Backup Path Patterns in Networks Telecommunication network owners and operators have for half a century been well aware of the potential loss of revenue if a major trunk is damaged, thus dependability at high cost has been implemented. A simple, effective and common dependability scheme is 1:1 protection with 100% capacity redundancy in the network. A growing number of applications in need of dependable connections with specific requirements to bandwidth and delay have started using the internet (which only provides best effort transport) as their base communication service. In this paper we adopt the 1:1 protection scheme and incorporate it as part of a routing system applicable for internet infrastructures. 100% capacity redundancy is no longer required. A distributed stochastic path finding (routing) algorithm based on Rubinstein's Cross Entropy method for combinatorial optimisation is presented. Early results from Monte Carlo simulations indeed indicate that the algorithm is capable of finding pairs of independent primary and backup paths satisfying specific bandwidth a constraints. Agents corradi01policy Policy Controlled Mobility The mobility of software components seems an interesting solution for the deployment of web services and applications in the Internet global infrastructure and also in mobile ad-hoc networks. The network infrastructure already supports several forms of code mobility to make possible to dynamically reconfigure bindings between code fragments and locations where they are to be executed. However, more work is still to be done to facilitate the specification and the enforcement of the mobility behaviour of software components. The traditional approach to embed the migration strategy into the component at design time can not suit the dynamicity of the new network scenarios. Mobile components should instead be enhanced with the possibility to adapt their mobility behaviour to evolving application and environment conditions and to react to unforeseen events. To reach this goal, the paper proposes the adoption of policy based systems to abstract away the specification of migration strategies from the component code. This approach permits to change the mobility behaviour of components without intervention on the component code. We have experienced the dynamicity and flexibility of the proposed approach in the framework obtained by integrating a policy-based management system in a mobile agent environment. Keywords: Mobility, Migration Policies, Adaptation Mobile Agents, Reconfiguration. 1 Agents 539969 User Behavior Analysis of Location Aware Search Engine Rapid growth of internet access from mobile users puts much importance on location specific information on the web. An unique web service called Mobile Info Search (MIS) from NTT Laboratories gathers the information and provide location aware search facilities. We performed association rule mining and sequence pattern mining against the access log which was accumulated at the MIS site in order to get some insight into the behavior of mobile users regarding the spatial information on the web. Detail web log mining process and the rules we derived are reported in this paper. IR vasconcelos00bayesian Bayesian Representations and Learning Mechanisms for Content-Based Image Retrieval We have previously introduced a Bayesian framework for content-based image retrieval (CBIR) that relies on a generative model for feature representation based on embedded mixtures. This is a truly generic image representation that can jointly model color and texture and has been shown to perform well across a broad spectrum of image databases. In this paper, we expand the Bayesian framework along two directions. First, we show that the formulation of CBIR as a problem of Bayesian inference leads to a natural criteria for evaluating local image similarity without requiring any image segmentation. This allows the practical implementation of retrieval systems where users can provide image regions, or objects, as queries. Region-based queries are significantly less ambiguous than queries based on entire images leading to significant improvements in retrieval precision. Second, we present a Bayesian learning algorithm that relies on belief propagation to integrate feedback provided by the... ML claypool99ounce An Ounce of Prevention is Worth A Pound of Cure: Formal Verification for Consistent Database Evolution Consistency of a database is as an important property that must be preserved at all times. In most OODB systems today, application code can directly access and alter both the data as well as the structure of the database. As a consequence application code can potentially violate the integrity of the database, in terms of the invariants of the data model, the user-specified application constraints, and even the referential integrity of the objects themselves. A common form of consistency management in most databases today is to encode constraints at the system level (e.g., foreign keys), or at the trigger based level (e.g., user constraints) and to perform transaction rollback on discovery of any violation of these constraints. However, for programs that alter the structure as well as the objects in a database, such as an extensible schema evolution program, roll-backs are expensive and add to the already astronomical cost of doing schema evolution. In this paper, pre-execution ... DB marini00specification Specification of Heterogeneous Agent Architectures . Agent-based software applications need to incorporate agents having heterogeneous architectures in order for each agent to optimally perform its task. HEMASL is a simple meta-language used to specify intelligent agents and multi-agent systems when different and heterogeneous agent architectures must be used. HEMASL specifications are based on an agent model that abstracts several existing agent architectures. The paper describes some of the features of the language, presents examples of its use and outlines its operational semantics. We argue that adding HEMASL to CaseLP, a specification and prototyping environment for MAS, can enhance its flexibility and usability. 1 Introduction Intelligent agents and multi-agent systems (MAS) are increasingly being acknowledged as the "new" modelling techniques to be used to engineer complex and distributed software applications [17, 9]. Agent-based software development is concerned with the realization of software applications modelled ... Agents bowling02multiagent Multiagent Learning Using a Variable Learning Rate Learning to act in a multiagent environment is a difficult problem since the normal definition of an optimal policy no longer applies. The optimal policy at any moment depends on the policies of the other agents. This creates a situation of learning a moving target. Previous learning algorithms have one of two shortcomings depending on their approach. They either converge to a policy that may not be optimal against the specific opponents ’ policies, or they may not converge at all. In this article we examine this learning problem in the framework of stochastic games. We look at a number of previous learning algorithms showing how they fail at one of the above criteria. We then contribute a new reinforcement learning technique using a variable learning rate to overcome these shortcomings. Specifically, we introduce the WoLF principle, “Win or Learn Fast, ” for varying the learning rate. We examine this technique theoretically, proving convergence in self-play on a restricted class of iterated matrix games. We also present empirical results on a variety of more general stochastic games, in situations of self-play and otherwise, demonstrating the wide applicability of this method. Key words: Multiagent learning, reinforcement learning, game theory 1 Agents 504171 Smart Playing Cards A Ubiquitous Computing Game Abstract. Recent technological advances allow for turning parts of our everyday environment into so–called smart environments. In this paper we present the “Smart Playing Cards ” application, a ubiquitous computing game that augments a classical card game with information–technological functionality, in contrast to developing new games around the abilities of available technology. Furthermore, we present the requirements such an application makes on a supporting software infrastructure for ubiquitous computing. HCI 232896 Query Optimization for Semistructured Data using Path Constraints in a Deterministic Data Model . Path constraints have been studied for semistructured data modeled as a rooted edge-labeled directed graph [4, 11--13]. In this model, the implication problems associated with many natural path constraints are undecidable [11, 13]. A variant of the graph model, called the deterministic data model , was recently proposed in [10]. In this model, data is represented as a graph with deterministic edge relations, i.e., the edges emanating from any node in the graph have distinct labels. This model is more appropriate for representing, e.g., ACeDB [27] databases and Web sites. This paper investigates path constraints for the deterministic data model. It demonstrates the application of path constraints to, among others, query optimization. Three classes of path constraints are considered: the language Pc introduced in [11], an extension of Pc , denoted by P w c , by including wildcards in path expressions, and a generalization of P w c , denoted by P c , by representing pa... DB hart99comparing Comparing Evolutionary Programs and Evolutionary Pattern Search Algorithms: A Drug Docking Application Evolutionary programs (EPs) and evolutionary pattern search algorithms (EPSAs) are two general classes of evolutionary methods for optimizing on continuous domains. The relative performance of these methods has been evaluated on standard global optimization test functions, and these results suggest that EPSAs more robustly converge to nearoptimal solutions than EPs. In this paper we evaluate the relative performance of EPSAs and EPs on a real-world application: flexible ligand binding in the Autodock docking software. We compare the performance of these methods on a suite of docking test problems. Our results confirm that EPSAs and EPs have comparable performance, and they suggest that EPSAs may be more robust on larger, more complex problems. 1 Introduction Evolutionary programs (EPs) and evolutionary pattern search algorithms (EPSAs) are two classes of evolutionary algorithms (EAs) that have been specifically developed for solving problems of the form min x2R n f(x): In particula... ML yao98making Making Use of Population Information in Evolutionary Artificial Neural Networks This paper is concerned with the simultaneous evolution of artificial neural network (ANN) architectures and weights. The current practice in evolving ANN's is to choose the best ANN in the last generation as the final result. This paper proposes a different approach to form the final result by combining all the individuals in the last generation in order to make best use of all the information contained in the whole population. This approach regards a population of ANN's as an ensemble and uses a combination method to integrate them. Although there has been some work on integrating ANN modules [2], [3], little has been done in evolutionary learning to make best use of its population information. Four linear combination methods have been investigated in this paper to illustrate our ideas. Three real-world data sets have been used in our experimental studies, which show that the recursive least-square (RLS) algorithm always produces an integrated system that outperforms the best individual. The results confirm that a population contains more information than a single individual. Evolutionary learning should exploit such information to improve generalization of learned systems. ML zambonelli01organisational Organisational Rules as an Abstraction for the Analysis and Design of Multi-Agent Systems Multi-agent systems... In this paper we introduce three additional organisational concepts - organisational rules, organisational structures, and organisational patterns - and discuss why we believe they are necessary for the complete specification of computational organisations. In particular, we focus on the concept of organisational rules and introduce a formalism, based on temporal logic, to specify them. This formalism is then used to drive the definition of the organisational structure and the identification of the organisational patterns. Finally, the paper sketches some guidelines for a methodology for agent-oriented systems based on our expanded set of organisational abstractions. Agents 119409 An Agent-Based Approach to the Construction of Floristic Digital Libraries This paper describes an agent-assisted approach to the construction of floristic digital libraries, which consist of very large botanical data repositories and related services. We propose an environment, termed Chrysalis, in which authors of plant morphologic descriptions can enter data into a digital library via a web-based editor. An agent that runs concurrently with the editor suggests potentially useful morphologic descriptions based on similar documents existing in the library. Benefits derived from the introduction of Chrysalis include reduced potential for errors and data inconsistencies, increased parallelism among descriptions, and considerable savings in the time regularly spent in visually checking for parallelism and manually editing data. KEYWORDS: agents, agent-based interfaces, floristic digital libraries, FNA, Chrysalis. INTRODUCTION Constructing the vast data repositories that will support knowledge-intensive activities in digital libraries poses problems of enormo... HCI kim99effective Effective Temporal Aggregation using Point-based Trees . Temporal databases introduce the concept of time into underlying data, and provide built-in facilities that allow users to store and retrieve time-varying data. The aggregation in temporal databases, that is, temporal aggregation is an extension of conventional aggregation on the domain and range of aggregates to include time concept. Temporal aggregation is important for various applications, but is very expensive. In this paper, we propose a new tree structure for temporal aggregation, called PA-tree, and aggregate processing method based on the PA-tree. We show that the time complexity of the proposed method is better than those of the existing methods. The time complexity of the proposed method is shown to be indeed the lower bound of the problem. We perform comparative experiments and show the performance advantage of our proposed method in practice. 1 Introduction While conventional database systems store the most recent snapshots of the real world, temporal datab... DB barros96business Business Suitability Principles for Workflow Modelling By incorporating aspects of coordination and collaboration, workflow implementations of information systems require a sound conceptualisation of business processing semantics. Traditionally, the success of conceptual modelling techniques has depended largely on the adequacy of conceptualisation, expressive power, comprehensibility and formal foundation. An equally important requirement, particularly with the increased conceptualisation of business aspects, is business suitability. In this paper, the focus is on the business suitability of workflow modelling for a commonly encountered class of (operational) business processing, e.g. those of insurance claims, bank loans and land conveyancing. A general assessment is first conducted on some integrated techniques characterising well-known paradigms - structured process modelling, object-oriented modelling, behavioural process modelling and business-oriented modelling. Through this, an insight into business suitability within the broader... HCI 539185 A Scalable and Ontology-Based P2P Infrastructure for Semantic Web Services Semantic Web Services are a promising combination of Semantic Web and Web service technology, aiming at providing means of automatically executing, discovering and composing semantically marked-up Web services. We envision peer-to-peer networks which allow for carrying out searches in real-time on permanently reconfiguring networks to be an ideal injastructure for deploying a network of Semantic Web Service providers. However, P2P networks evolving in an unorganized manner suffer jom serious scalability problems, limiting the number of nodes in the network, creating network overload and pushing search times to unacceptable limits. We address these problems by imposing a deterministic shape on P2P networks: We propose a graph topology which allows for very efficient broadcast and search, and we provide an efficient topology construction and maintenance algorithm which, crucial to symmetric peer-to-peer networks, does neither require a central server nor super nodes in the network. We show how our scheme can be made even more efficient by using a globally known ontology to determine the organization of peers in the graph topology, allowing for efficient concept-based search. HCI armstrong99boticelli Boticelli: A Single-Camera Mobile Robot Using New Approaches to Range Data Fusion, World Modeling, and Navigation Planning ID: A083 Abstract Boticelli is a mobile robot, designed and built for testing new approaches in stereo vision, world modeling, data fusion, map extraction, reinforcement learning and navigation planning. A single camera is used to capture depth information by taking advantage of camera movements. The main thrust of the new approaches is to replace well-known techniques, that depend upon grids of points in space with techniques that use continuous, piecewise linear functions. These techniques scale well to large, complex environments. Keywords: mobile agents, mapping and exploration, reinforcement learning Acknowledgements This research was supported mainly by the Defence Research Establishment Suffield (contract W7702-6R594 /001 to Dendronic Decisions Limited). We are very grateful to the Scientific Authority Dr. Simon Barton for his guidance. Travel was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC). Help of Kyle Palmer in designing the multi... AI 114630 Response Generation in Collaborative Negotiation In collaborative planning activities, since the agents are autonomous and heterogeneous, it is inevitable that conflicts arise in their beliefs during the planning process. In cases where such conflicts are relevant to the task at hand, the agents should engage in collaborative negotiation as an attempt to square away the discrepancies in their beliefs. This paper presents a computational strategy for detecting conflicts regarding proposed beliefs and for engaging in collaborative negotiation to resolve the conflicts that warrant resolution. Our model is capable of selecting the most effective aspect to address in its pursuit of conflict resolution in cases where multiple conflicts arise, and of selecting appropriate evidence to justify the need for such modification. Furthermore, by capturing the negotiation process in a recursive Propose-Evaluate-Modify cycle of actions, our model can successfully handle embedded negotiation subdialogues. 1 Introduction In collaborative consultat... Agents rosenthal00view View Security as the Basis for Data Warehouse Security Access . permissions in a data warehouse are currently managed in a separate world from the sources' policies. The consequences are inconsistencies, slow response to change, and wasted administrative work. We present a different approach, which treats the sources' exported tables and the warehouse as part of the same distributed database. Our main result is a way to control derived products by extending SQL grants rather than creating entirely new mechanisms. We provide a powerful, sound inference theory that derives permissions on warehouse tables (both materialized and virtual), making the system easier to administer and its applications more robust. We also propose a new permission construct suitable for views that filter data from mutually-suspicious parties. 1 Introduction A key challenge for data warehouse security is how to manage the entire system coherently -- from sources and their export tables, to warehouse stored tables (conventional and cubes) and vi... DB mazumdar99achieving Achieving Consistency in Mobile Databases through Localization in PRO-MOTION There is great need and potential for traditional transaction support in a mobile computing environment. However, owing to the inherent limitations of mobile computing, we need to augment the well-developed techniques of Database Management Systems with new approaches. In this paper, we focus on the challenge of assuring data consistency. Our approach of localization is to reformulate global constraints so as to enhance the autonomy of the mobile hosts. We show how this approach unifies techniques of maintaining replicated data with methods of enforcing polynomial inequalities. We also discuss how localization can be implemented in PRO-MOTION, a flexible infrastructure for transaction processing in a mobile environment. 1. Introduction Thanks to the relentless advances in semiconductors, the number of users with mobile computers (we will refer to these machines as mobile hosts or MHs) continues to increase. These users have discovered that exciting developments in wireless technology... DB claypool99optimizatizing Optimizatizing the Performance of Schema Evolution Sequences More than ever before schema transformation is a prevalent problem that needs to be addressed to accomplish for example the migration of legacy systems to the newer OODB systems, the generation of structured web pages from data in database systems, or the integration of systems with different native data models. Such schema transformations are typically composed of a sequence of schema evolution operations. The execution of such sequences can be very timeintensive, possibly requiring many hours or even days and thus effectively making the database unavailable for unacceptable time spans. While researchers have looked at the deferred execution approach for schema evolution in an effort to improve availability of the system, to the best of our knowledge ours is the first effort to provide a direct optimization strategy for a sequence of changes. In this paper, we propose heuristics for the iterative elimination and cancellation of schema evolution primitives as well as for the merging of... DB kosala00web Web Mining Research: A Survey With the huge amount of information available online, the World Wide Web is a fertile area for data mining research. The Web mining research is at the cross road of research from several research communities, such as database, information retrieval, and within AI, especially the sub-areas of machine learning and natural language processing. However, there is a lot of confusions when comparing research efforts from different point of views. In this paper, we survey the research in the area of Web mining, point out some confusions regarded the usage of the term Web mining and suggest three Web mining categories. Then we situate some of the research with respect to these three categories. We also explore the connection between the Web mining categories and the related agent paradigm. For the survey, we focus on representation issues, on the process, on the learning algorithm, and on the application of the recent works as the criteria. We conclude the paper with some research issues. IR wilensky00digital Digital Library Resources as a Basis for Collaborative Work The creation of large, networked, digital document resources has greatly facilitated information access and dissemination. We suggest that such resources can further enhance how we work with information, namely, that they can provide a substrate that supports collaborative work. We focus on one form of collaboration, annotation, by which we mean any of an open-ended number of creative document manipulations which are useful to record and to share with others. Widespread digital document dissemination required technological enablers, such as web clients and servers. The resulting infrastructure is one in which information may be widely shared by individuals across administrative boundaries. To achieve the same ubiquitous availability for annotation requires providing support for spontaneous collaboration, that is, for collaboration across administrative boundaries without significant prior agreements. Annotation is not more commonplace, we suggest, because the technological needs of sp... IR laviola99msvt MSVT: A Virtual Reality-Based Multimodal Scientific Visualization Tool Recent approaches to providing users with more natural methods of interacting with virtual environment applications have shown that more than one mode of input can be both beneficial and intuitive as a communication medium between humans and computer applications. Although there are many different modes that could be used in these applications, hand gestures and speech appear to be two of the most logical since users will typically be in environments that will have them immersed in a virtual world with limited access to traditional input devices such as the keyboard or mouse. In this paper, we describe a prototype application, MSVT (Multimodal Scientific Visualization Tool), for visualizing fluid flow around a dataset. MSVT uses a multimodal interface which combines whole-hand and voice input to allow users to visualize and interact with the dataset in a natural manner. A discussion of the various interaction techniques, and the results of an informal user evaluation are presented. KE... HCI vardy99wristcam The WristCam as Input Device We show how images of a user's hand from a video camera attached to the underside of the wrist can be processed to yield finger movement information. Discrete (and discreet) movements of the fingers away from a rest position are translated into a small set of base symbols. These are interpreted as input to a wearable computer, providing unobtrusive control. HCI bradford98pruning Pruning Decision Trees with Misclassification Costs . We describe an experimental study of pruning methods for decision tree classifiers when the goal is minimizing loss rather than error. In addition to two common methods for error minimization, CART's cost-complexity pruning and C4.5's error-based pruning, we study the extension of cost-complexity pruning to loss and one pruning variant based on the Laplace correction. We perform an empirical comparison of these methods and evaluate them with respect to loss. We found that applying the Laplace correction to estimate the probability distributions at the leaves was beneficial to all pruning methods. Unlike in error minimization, and somewhat surprisingly, performing no pruning led to results that were on par with other methods in terms of the evaluation criteria. The main advantage of pruning was in the reduction of the decision tree size, sometimes by a factor of ten. While no method dominated others on all datasets, even for the same domain different pruning mechanisms are better for ... ML wicke99compiling Compiling for Fast State Capture of Mobile Agents Saving, transporting, and restoring the state of a mobile agent is one of the main problems in implementing a mobile agents system. We present an approach, implemented as part of our Messengers system, that represents a trade-off between the unrestricted use of pointers and the ability to perform fully transparent state capture. When writing the code for an agent, the programmer has a choice between two types of functions. C functions are fully general and may use unrestricted pointers, but they are not allowed to invoke any migration commands. Messengers functions may cause migration but their use of pointers is restricted to only a special type of a dynamic array structure. Under these restrictions, the local variables, the program counter, and the calling stack of an agent can all be made machine-independent and can be captured/restored transparently during migration. 1 Introduction Saving, transporting, and restoring the state of a mobile agent is one of the main problem in implem... Agents krum02speech Speech and Gesture Multimodal Control of a Whole Earth 3D Visualization Environment A growing body of research shows several advantages to multimodal interfaces including increased expressiveness, flexibility, and user freedom. This paper investigates the design of such an interface that integrates speech and hand gestures. The interface has the additional property of operating relative to the user and can be used while the user is in motion or stands at a distance from the computer display. The paper then describes an implementation of the multimodal interface for a whole earth 3D visualization environment which presents navigation interface challenges due to the large magnitude of scale and extended spaces that is available. The characteristics of the multimodal interface are examined, such as speed, recognizability of gestures, ease and accuracy of use, and learnability under likely conditions of use. This implementation shows that such a multimodal interface can be e#ective in a real environment and sets some parameters for the design and use of such interfaces. HCI hodkinson99decidable Decidable Fragments of First-Order Temporal Logics In this paper, we introduce a new fragment of the first-order temporal language, called the monodic fragment, in which all formulas beginning with a temporal operator (Since or Until) have at most one free variable. We show that the satisfiability problem for monodic formulas in various linear time structures can be reduced to the satisfiability problem for a certain fragment of classical first-order logic. This reduction is then used to single out a number of decidable fragments of first-order temporal logics and of two-sorted first-order logics in which one sort is intended for temporal reasoning. Besides standard first-order time structures, we consider also those that have only finite first-order domains, and extend the results mentioned above to temporal logics of finite domains. We prove decidability in three different ways: using decidability of monadic second-order logic over the intended flows of time, by an explicit analysis of structures with natural numbers time, and by a composition method that builds a model from pieces in finitely many steps. 1 DB 455651 Text Database Selection for Longer Queries A metasearch engine is a system that supports unified access to multiple local search engines. One of the main challenges in building a large-scale metasearch engine is to solve the database (search engine) selection problem, which is to efficiently and accurately determine a small number of potentially useful local search engines to invoke for each user query. For the database of each search engine, a representative which indicates approximately the contents of the database is created in advance to enable database selection. The representatives of all databases can be integrated into a single representative to make the selection process more scalable. While an integrated representative with high scalability has just been proposed and has been found to be effective for short queries, its effectiveness for longer queries is significantly lower. In the Internet environment, most queries initially submitted by users are short queries. However, it has been found that better search effectiveness can often be achieved when additional terms are added to the initial queries through query expansion or relevance feedback. The resulting queries are usually longer than the initial queries. In this paper, we propose a new method to construct database representatives and to decide which databases to select for longer queries. Experimental results are given to compare the performance of the new method with that of a previous method. IR ram90incremental Incremental Learning of Explanation Patterns and their Indices This paper describes how a reasoner can improve its understanding of an incompletely understood domain through the application of what it already knows to novel problems in that domain. Recent work in AI has dealt with the issue of using past explanations stored in the reasoner's memory to understand novel situations. However, this process assumes that past explanations are well understood and provide good "lessons" to be used for future situations. This assumption is usually false when one is learning about a novel domain, since situations encountered previously in this domain might not have been understood completely. Instead, it is reasonable to assume that the reasoner would have gaps in its knowledge base. By reasoning about a new situation, the reasoner should be able to fill in these gaps as new information came in, reorganize its explanations in memory, and gradually evolve a better understanding of its domain. We present a story understanding program that retrieves past explan... ML omlin98equivalence Equivalence in Knowledge Representation: Automata, Recurrent Neural Networks, and Dynamical Fuzzy Systems Neurofuzzy systems-the combination of artificial neural networks with fuzzy logic-have become useful in many application domains. However, conventional neurofuzzy models usually need enhanced representation power for applications that require context and state (e.g., speech, time series prediction, control). Some of these applications can be readily modeled as finite state automata. Previously, it was proved that deterministic finite state automata (DFA) can be synthesized by or mapped into recurrent neural networks by directly programming the DFA structure into the weights of the neural network. Based on those results, a synthesis method is proposed for mapping fuzzy finite state automata (FFA) into recurrent neural networks. Furthermore, this mapping is suitable for direct implementation in very large scale integration (VLSI), i.e., the encoding of FFA as a generalization of the encoding of DFA in VLSI systems. The synthesis method requires FFA to undergo a transformation prior to being mapped into recurrent networks. The neurons are provided with an enriched functionality in order to accommodate a fuzzy representation of FFA states. This enriched neuron functionality also permits fuzzy parameters of FFA to be directly represented as parameters of the neural network. We also prove the stability of fuzzy finite state dynamics of the constructed neural networks for finite values of network weight and, through simulations, give empirical validation of the proofs. Hence, we prove various knowledge equivalence representations between neural and fuzzy systems and models of automata. ML 32215 Using an Explicit Teamwork Model and Learning in RoboCup: An Extended Abstract Stacy Marsella, Jafar Adibi, Yaser Al-Onaizan, Ali Erdem, Randall Hill Gal A. Kaminka, Zhun Qiu, Milind Tambe Information Sciences Institute and Computer Science Department University of Southern California 4676 Admiralty Way, Marina del Rey, CA 90292, USA robocup-sim@isi.edu 1 Introduction The RoboCup research initiative has established synthetic and robotic soccer as testbeds for pursuing research challenges in Artificial Intelligence and robotics. This extended abstract focuses on teamwork and learning, two of the multiagent research challenges highlighted in RoboCup. To address the challenge of teamwork, we discuss the use of a domain-independent explicit model of teamwork, and an explicit representation of team plans and goals. We also discuss the application of agent learning in RoboCup. The vehicle for our research investigations in RoboCup is ISIS (ISI Synthetic), a team of synthetic soccer-players that successfully participated in the simulation league of RoboCup'97, by win... ML 460667 Hardware/Software Implementation for Localization and Classification Systems Pattern localization and classification are CPU time intensive being normally implemented in software, however with lower performance than custom implementations. Custom implementation in hardware (ASIC) allows real-time processing, having higher cost and time-to-market than software implementation. We present an alternative that represents a good trade-off between performance and cost. This paper presents initially some systems dealing with object localization and classification, analyzing the performance and implementation of each work. After we propose a system for localization and classification of shapes using reconfigurable devices (FPGA) and a signal processor (DSP) available in a flexible codesign platform. The system will be described using C and VHDL languages, for the software and hardware parts respectively, and has been implemented in an APTIX prototyping platform. 1. ML chiu99meta A Meta Modeling Approach to Workflow Management Systems Supporting Exception Handling Workflow Management Systems (WFMSs) facilitate the definition of structure and decomposition of business processes and assists in management of coordinating, scheduling, executing and monitoring of such activities. Most of the current WFMSs are built on traditional relational database systems and/or using an objectoriented database system for storing the definition and run time data about the workflows. However, a WFMS requires advanced modeling functionalities to support adaptive features, such as on-line exception handling. This article describes our advanced meta-modeling approach using various enabling technologies (such as object orientation, roles, rules, active capabilities) supported by an integrated environment, the ADOME, as a solid basis for a flexible WFMS involving dynamic match making, migrating workflows and exception handling. Copyright 1999 Elsevier Science Ltd Key words: Meta-modeling, Object-Orientation, Workflow Management, Match-Making, Exception Handling, Workflo... DB touzet01side SIDE Surfer: a Spontaneous Information Discovery and Exchange System Development of wireless communications enables the rise of networking applications in embedded systems. Web interactions, which are the most spread, are nowadays available on wireless PDAs. Moreover, we can observe a development of ubiquitous computing. Based on this concept, many works aim to consider user's context as part of the parameters of the applications. The context notion can include the user's location, his social activity . . . Taking part from emerging technologies enabling short range and direct wireless communications (which allow to define a proximity context), the aim of our study is to design a new kind of application, extending the Web paradigm: spontaneous and proximate Web interactions. HCI beigi97metaseek MetaSEEk: A Content-Based Meta-Search Engine for Images Search engines are the most powerful resources for finding information on the rapidly expanding World Wide Web (WWW). Finding the desired search engines and learning how to use them, however, can be very time consuming. The integration of such search tools enables the users to access information across the world in a transparent and efficient manner. These systems are called meta-search engines. The recent emergence of visual information retrieval (VIR) search engines on the web is leading to the same efficiency problem. This paper describes and evaluates MetaSEEk, a content-based meta-search engine used for finding images on the Web based on their visual information. MetaSEEk is designed to intelligently select and interface with multiple on-line image search engines by ranking their performance for different classes of user queries. User feedback is also integrated in the ranking refinement. We compare MetaSEEk with a base line version of meta-search engine, which does not use the past performance of the different search engines in recommending target search engines for future queries. IR langley95acquisition Acquisition of Place Knowledge Through Case-Based Learning In this paper we define the task of place learning and describe one approach to this problem. The framework represents distinct places as evidence grids, a probabilistic description of occupancy. Place recognition relies on case-based classification, augmented by a registration process to correct for translations. The learning mechanism is also similar to that in case-based systems, involving the simple storage of inferred evidence grids. Experimental studies with physical and simulated robots suggest that this approach improves place recognition with experience, that it can handle significant sensor noise, that it benefits from improved quality in stored cases, and that it scales well to environments with many distinct places. Previous researchers have studied evidence grids and place learning, but they have not combined these two powerful concepts, nor have they used the experimental methods of machine learning to evaluate their methods' abilities. Keywords: place acquisition, case-b... ML joho02hierarchical Hierarchical Presentation of Expansion Terms Different presentations of candidate expansion terms have not been fully explored in interactive query expansion (IQE). Most existing systems that offer an IQE facility use a list form of presentation. This paper examines an hierarchical presentation of the expansion terms which are automatically generated from a set of retrieved documents, organised in a general to specific manner, and visualised by cascade menus. To evaluate the effectiveness of the presentation, a user test was carried out to compare the hierarchical form with the conventional list form. This shows that users of the hierarchy can complete the expansion task in less time and with fewer terms over those using the lists. Relations between initial query terms and selected expansion terms were also investigated. Keywords Information retrieval, interactive query expansion, concept hierarchies 1. HCI kadobayashi98seamless Seamless Guidance by Personal Agent in Virtual Space Based on User Interaction in Real World This paper describes a personal agent that guides a visitor through a virtual space by using context, i.e., visitor interaction, in the real world. #Context-aware# guidance in the real world has become an increasingly important research demand since the advent of small and powerful palmtop computers, or PDAs (Personal Digital Assistants), that can obtain useful information about objects in the real world based on context (e.g., current location and visit history) in the real world. We have tried to extend this context-aware guidance into a virtual environment by using context not from the virtual environment but from the real world; such seamless guidance allows visitors to concentrate their attention on the understanding of actual elements. This paper reports our experiment on seamless guidance by a personal agent in a virtual ancient village using real-world context. This was presented at an open house exhibition in our research laboratories. 1 Introduction We have proposed a notion... AI hunter00merging Merging Potentially Inconsistent Items of Structured Text Structured text is a general concept that is implicit in a variety of approaches to handling information. Syntactically, an item of structured text is a number of grammatically simple phrases together with a semantic label for each phrase. Items of structured text may be nested within larger items of structured text. The semantic labels in a structured text are meant to parameterize a stereotypical situation, and so a particular item of structured text is an instance of that stereotypical situation. Much information is potentially available as structured text including tagged text in XML, text in relational and object-oriented databases, and the output from information extraction systems in the form of instantiated templates. In this paper, we formalize the concept of structured text, and then focus on how we can identify inconsistency in the logical representation of items of structured text. We then present a new framework for merging logical theories that can be employed t... DB cabri00mars MARS: a Programmable Coordination Architecture for Mobile Agents Mobile agents represent a promising technology for the development of Internet applications. However, mobile computational entities introduce peculiar problems w.r.t. the coordination of the application components. The paper outlines the advantages of Linda-like coordination models, and shows how a programmable coordination model based on reactive tuple spaces can provide further desirable features for Internet applications based on mobile agents. Accordingly, the paper presents the design and the implementation of the MARS coordination architecture for Java-based mobile agents. MARS defines Linda-like tuple spaces, which can be programmed to react with specific actions to the accesses made by mobile agents. Agents 534078 Text-Based Content Search and Retrieval in ad hoc P2P Communities We consider the problem of content search and retrieval in peer-to-peer (P2P) communities. P2P computing is a potentially powerful model for information sharing between ad hoc groups' of users because of its' low cost of entry and natural model for resource scaling with community size. As P2P communities grow in size, however, locating information distributed across the large number of peers becomes problematic. We present a distributed text-based content search and retrieval algorithm to address this' problem. Our algorithm is' based on a state-of-the-art text-based document ranking algorithm: the vector-space model instantiated with the TFxlDF ranking rule. A naive application of TFxlDF wouM require each peer in a community to collect an inverted index of the entire community. This' is' costly both in terms of bandwidth and storage. Instea & we show how TFxlDF can be approximated given compact summaries of peers' local inverted indexes. We make three contributions: (a) we show how the TFxlDF rule can be adapted to use the index summaries, (b) we provide a heuristic for adaptively determining the set of peers that shouM be contacted for a query, and (c) we show that our algorithm tracks' TFxlDF's performance very closely, regardless of how documents' are distributed throughout the community. Furthermore, our algorithm preserves the main flavor of TFxlDF by retrieving close to the same set of documents for any given query. IR 84828 Projective Rotations applied to a Pan-Tilt Stereo Head A non-metric pan-tilt stereo-head consists of a weakly calibrated stereo rig mounted on a pan-tilt mechanism. It is called non-metric since neither the kinematics of the mechanism, nor camera calibration are required. The Lie group of "projective rotations"- homographies of projective space corresponding to pure rotations -- is an original formalism to model the geometry of such a pan-tilt system. A Rodrigues alike formula as well as a minimal parameterization of projective rotations are introduced. Based on this, the practical part devises a numerical optimization technique for accurately estimating projective rotations from point correspondences, only. This procedure recovers sufficient geometry to operate the system. The experiments validate and evaluate the proposed approach on real image data. They show the weak calibration, image prediction, and homing of a non-metric pan-tilt head. 1 Introduction One of the most useful sensors in computer vision is a pan and tilt stereo head. ... ML minar99cooperating Cooperating Mobile Agents for Dynamic Network Routing this paper we present a contrasting model, a dynamic, wireless, peer to peer network with routing tasks performed in a decentralized and distributed fashion by mobile software agents that cooperate to accumulate and distribute connectivity information. Our agents determine system topology by exploring the network, then store this information in the nodes on the network. Other agents use this stored information to derive multi-hop routes across the network. We study these algorithms in simulation as an example of using populations of mobile agents to manage networks Agents hellerstein98optimization Optimization Techniques For Queries with Expensive Methods Object-Relational database management systems allow knowledgeable users to de ne new data types, as well as new methods (operators) for the types. This exibility produces an attendant complexity, which must be handled in new ways for an Object-Relational database management system to be e cient. In this paper we study techniques for optimizing queries that contain time-consuming methods. The focus of traditional query optimizers has been on the choice of join methods and orders; selections have been handled by \pushdown " rules. These rules apply selections in an arbitrary order before as many joins as possible, using the assumption that selection takes no time. However, users of Object-Relational systems can embed complex methods in selections. Thus selections may take signi cant amounts of time, and the query optimization model must be enhanced. In this paper, we carefully de ne a query cost framework that incorporates both selectivity and cost estimates for selections. We develop an algorithm called Predicate Migration, and prove that it produces optimal plans for queries with expensive methods. We then describe our implementation of Predicate Migration in the commercial Object-Relational database management system Illustra, and discuss practical issues that a ect our earlier assumptions. We compare Predicate Migration to a variety of simpler optimization techniques, and demonstrate that Predicate Migration is the best general solution to date. The alternative techniques we presentmaybe useful for constrained workloads. DB braumandl00functional Functional Join Processing . Inter-object references are one of the key concepts of object-relational and object-oriented database systems. In this work, we investigate alternative techniques to implement inter-object references and make the best use of them in query processing, i.e., in evaluating functional joins. We will give a comprehensive overview and performance evaluation of all known techniques for simple (singlevalued) as well as multi-valued functional joins. Furthermore, we will describe special order-preserving functionaljoin techniques that are particularly attractive for decision support queries that require ordered results. While most of the presentation of this paper is focused on object-relational and object-oriented database systems, some of the results can also be applied to plain relational databases because index nested-loop joins along key/foreign-key relationships, as they are frequently found in relational databases, are just one particular way to execute a functional join. Key words: O... DB michaud99managing Managing Robot Autonomy and Interactivity Using Motives and Visual Communication An autonomous mobile robot operating in everyday life conditions will have to face a huge variety of situations and to interact with other agents (living or artificial). Such a robot needs flexible and robust methods for managing its goals and for adapting its control mechanisms to face the contingencies of the world. It also needs to communicate with others in order to get useful information about the world. This paper describes an approach based on a general architecture and on internal variables called `motives' to manage the goals of an autonomous robot. These variables are also used as a basis for communication using a visual communication system. Experiments using a vision- and sonar-based Pioneer I robot, equipped with a visual signaling device, are presented. 1 Introduction Designing an autonomous mobile robot to operate in unmodified environments, i.e., environments that have not been specifically engineered for the robot, is a very challenging problems. Dynamic and unpredic... AI chang99approximate Approximate Query Translation across Heterogeneous Information Sources In this paper we present a mechanism for approximately translating Boolean query constraints across heterogeneous information sources. Achieving the best translation is challenging because sources support different constraints for formulating queries, and often these constraints cannot be precisely translated. For instance, a query [score ? 8] might be "perfectly" translated as [rating ? 0.8] at some site, but can only be approximated as [grade = A] at another. Unlike other work, our general framework adopts a customizable "closeness" metric for the translation that combines both precision and recall. Our results show that for query translation we need to handle interdependencies among both query conjuncts as well as disjuncts. As the basis, we identify the essential requirements of a rule system for users to encode the mappings for atomic semantic units. Our algorithm then translates complex queries by rewriting them in terms of the semantic units. We show that, un... DB muller01cube The Cub-e, a Novel Virtual 3D Display Device We have designed, and are in the process of building, a visualisation device, the Cub-e. The Cub-e consists of six TFT screens, arranged in a perspex cube, with a StrongARM processor and batteries inside. It is a multipurpose device with applications including teleconferencing, interaction with virtual worlds, and games. 1 HCI schweighofer01automatic Automatic Text Representation, Classification and Labeling in European Law The huge text archives and retrieval systems of legal information have not achieved yet the representation in the wellknown subject-oriented structure of legal commentaries. Content-based classification and text analysis remains a high priority research topic. In the joint KONTERM, SOM and LabelSOM projects, learning techniques of neural networks are used to achieve similar high compression rates of classification and analysis like in manual legal indexing. The produced maps of legal text corpora cluster related documents in units that are described with automatically selected descriptors. Extensive tests with text corpora in European case law have shown the feasibility of this approach. Classification and labeling proved very helpful for legal research. The Growing Hierarchical Self-Organizing Map represents very interesting generalities and specialties of legal text corpora. The segmentation into document parts improved very much the quality of labeling. The next challenge would be a change from tfxidf vector representation to a modified vector representation taking into account thesauri or ontologies considering learned properties of legal text corpora. IR schuldt99concurrency Concurrency Control and Recovery in Transactional Process Management The classical theory of transaction management contains two different aspects, namely concurrency control and recovery, which ensure serializability and atomicity of transaction executions, respectively. Although concurrency control and recovery are not independent of each other, the criteria for these two aspects were developed orthogonally and as a result, in most cases these criteria are incompatible with each other. Recently a unified theory of concurrency control and recovery for databases with read and write operations has been introduced in [SWY93, AVA + 94] that allows reasoning about serializability and atomicity within the same framework. In [SWY93, AVA + 94] a class of schedules was introduced (called prefix reducible), which guarantees both serializability and atomicity in a failure prone environment with read/write operations. Several protocols were developed to generate such schedules by a database concurrency control mechanism. We present here a unified transaction ... DB klein91supporting Supporting Conflict Resolution in Cooperative Design Systems Complex modern-day artifacts are designed cooperatively by groups of experts, each with their own areas of expertise. The interaction of such experts inevitably involves conflict. This paper presents an implemented computational model, based on studies of human cooperative design, for supporting the resolution of such conflicts. This model is based centrally on the insights that general conflict resolution expertise exists separately from domain-level design expertise, and that this expertise can be instantiated in the context of particular conflicts into specific advice for resolving those conflicts. Conflict resolution expertise consists of a taxonomy of design conflict classes in addition to associated general advice suitable for resolving conflicts in these classes. The abstract nature of conflict resolution expertise makes it applicable to a wide variety of design domains. This paper describes this conflict resolution model and provides examples of its operation from an implemente... Agents 29551 CMUnited-97: RoboCup-97 Small-Robot World Champion Team Robotic soccer is a challenging research domain which involves multiple agents that need to collaborate in an adversarial environment to achieve specificobjectives. In this paper, we describe CMUnited, the team of small robotic agents that we developed to enter the RoboCup-97 competition. We designed and built the robotic agents, devised the appropriate vision algorithm, and developed and implemented algorithms for strategic collaboration between the robots in an uncertain and dynamic environment. The robots can organize themselves in formations, hold specific roles, and pursue their goals. In game situations, they have demonstrated their collaborative behaviors on multiple occasions. We present an overview of the vision processing algorithm which successfully tracks multiple moving objects and predicts trajectories. The paper then focusses on the agent behaviors ranging from low-level individual behaviors to coordinated, strategic team behaviors. CMUnited won the RoboCup-97 small-robot competition at IJCAI-97 in Nagoya, Japan. AI fang98computing Computing Iceberg Queries Efficiently Many applications compute aggregate functions (such as COUNT, SUM) over an attribute (or set of attributes) to find aggregate values above some specified threshold. We call such queries iceberg queries because the number of above-threshold results is often very small (the tip of an iceberg), relative to the large amount of input data (the iceberg). Such iceberg queries are common in many applications, including data warehousing, information-retrieval, market basket analysis in data mining, clustering and copy detection. We propose efficient algorithms to evaluate iceberg queries using very little memory and significantly fewer passes over data, as compared to current techniques that use sorting or hashing. We present an experimental case study using over three gigabytes of Web data to illustrate the savings obtained by our algorithms. 1 Introduction In this paper we develop efficient execution strategies for an important class of queries that we call iceberg queries. An iceberg query... IR michail99assessing Assessing Software Libraries by Browsing Similar Classes, Functions, and Relationships Comparing and contrasting a set of software libraries is useful for reuse related activities such as selecting a library from among several candidates or porting an application from one library to another. The current state of the art in assessing libraries relies on qualitative methods. To reduce costs and/or assess a large collection of libraries, automation is necessary. Although there are tools that help a developer examine an individual library in terms of architecture, style, etc., we know of no tools that help the developer directly compare several libraries. With existing tools, the user must manually integrate the knowledge learned about each library. Automation to help developers directly compare and contrast libraries requires matching of similar components (such as classes and functions) across libraries. This is different than the traditional component retrieval problem in which components are returned that best match a user's query. Rather, we need to find those component... IR 538013 Self-Organization in Ad Hoc Sensor Networks: An Empirical Study Research in classifying and recognizing complex concepts has been directing its focus increasingly on distributed sensing using a large amount of sensors. The colossal amount of sensor data often obstructs traditional algorithms in centralized approaches, where all sensor data is directed to one central location to be processed. Spreading the processing of sensor data over the network seems to be a promising option, but distributed algorithms are harder to inspect and evaluate. Using self-sufficient sensor boards with short-range wireless communication capabilities, we are exploring approaches to achieve an emerging distributed perception of the sensed environment in realtime through clustering. Experiments in both simulation and real-world platforms indicate that this is a valid methodology, being especially promising for computation on many units with limited resources. HCI ganti98clustering Clustering Large Datasets in Arbitrary Metric Spaces Clustering partitions a collection of objects into groups called clusters, such that similar objects fall into the same group. Similarity between objects is defined by a distance function satisfying the triangle inequality; this distance function along with the collection of objects describes a distance space. In a distance space, the only operation possible on data objects is the computation of distance between them. All scalable algorithms in the literature assume a special type of distance space, namely a k-dimensional vector space, which allows vector operations on objects. We present two scalable algorithms designed for clustering very large datasets in distance spaces. Our first algorithm BUBBLE is, to our knowledge, the first scalable clustering algorithm for data in a distance space. Our second algorithm BUBBLE-FM improves upon BUBBLE by reducing the number of calls to the distance function, which may be computationally very expensive. Both algorithms make only a single scan ov... ML grumbach00manipulating Manipulating Interpolated Data is Easier than You Thought Data defined by interpolation is frequently found in new applications involving geographical entities, moving objects, or spatiotemporal data. These data lead to potentially infinite collections of items, (e.g., the elevation of any point in a map), whose definitions are based on the association of a collection of samples with an interpolation function. The naive manipulation of the data through direct access to both the samples and the interpolation functions leads to cumbersome or inaccurate queries. It is desirable to hide the samples and the interpolation functions from the logical level, while their manipulation is performed automatically. We propose to model such data using infinite relations (e.g., the map with elevation yields an infinite ternary relation) which can be manipulated through standard relational query languages (e.g., SQL), with no mention of the interpolated definition. The clear separation between logical and physical levels ensures the accu... DB cheverst00developing Developing a Context-aware Electronic Tourist Guide: Some Issues and Experiences In this paper, we describe our experiences of developing and evaluating GUIDE, an intelligent electronic tourist guide. The GUIDE system has been built to overcome many of the limitations of the traditional information and navigation tools available to city visitors. For example, group-based tours are inherently inflexible with fixed starting times and fixed durations and (like most guidebooks) are constrained by the need to satisfy the interests of the majority rather than the specific interests of individuals. Following a period of requirements capture, involving experts in the field of tourism, we developed and installed a system for use by visitors to Lancaster. The system combines mobile computing technologies with a wireless infrastructure to present city visitors with information tailored to both their personal and environmental contexts. In this paper we present an evaluation of GUIDE, focusing on the quality of the visitors experience when using the system. Keywords Mobile c... HCI 51542 Reasoning with Concrete Domains Description logics are knowledge representation and reasoning formalisms which represent conceptual knowledge on an abstract logical level. Concrete domains are a theoretically well-founded approach to the integration of description logic reasoning with reasoning about concrete objects such as numbers, time intervals or spatial regions. In this paper, the complexity of combined reasoning with description logics and concrete domains is investigated. We extend ALC(D), which is the basic description logic for reasoning with concrete domains, by the operators "feature agreement" and "feature disagreement". For the extended logic, called ALCF(D), an algorithm for deciding the ABox consistency problem is devised. The strategy employed by this algorithm is vital for the efficient implementation of reasoners for description logics incorporating concrete domains. Based on the algorithm, it is proved that the standard reasoning problems for both logics ALC(D) and ALCF(D) are PSpace-co... AI amiri00highly Highly Concurrent Shared Storage . Shared storage arrays enable thousands of storage devices to be shared and directly accessed by end hosts over switched system-area networks, promising databases and filesystems highly scalable, reliable storage. In such systems, hosts perform access tasks (read and write) and management tasks (migration and reconstruction of data on failed devices.) Each task translates into multiple phases of low-level device I/Os, so that concurrent host tasks can span multiple shared devices and access overlapping ranges potentially leading to inconsistencies for redundancy codes and for data read by end hosts. Highly scalable concurrency control and recovery protocols are required to coordinate on-line storage management and access tasks. While expressing storage-level tasks as ACID transactions ensures proper concurrency control and recovery, such an approach imposes high performance overhead, results in replication of work and does not exploit the available knowledge about storage le... DB lee01collaborative Collaborative Learning for Recommender Systems Recommender systems use ratings from users on items such as movies and music for the purpose of predicting the user preferences on items that have not been rated. Predictions are normally done by using the ratings of other users of the system, by learning the user preference as a function of the features of the items or by a combination of both these methods. In this paper, we pose the problem as one of collaboratively learning of preference functions by multiple users of the recommender system. We study several mixture models for this task. We show, via theoretical analyses and experiments on a movie rating database, how the models can be designed to overcome common problems in recommender systems including the new user problem, the recurring startup problem, the sparse rating problem and the scaling problem. 1. IR 87014 Dynamic Service Matchmaking Among Agents in Open Information Environments Introduction The amount of services and deployed software agents in the most famous offspring of the Internet, the World Wide Web, is exponentially increasing. In addition, the Internet is an open environment, where information sources, communication links and agents themselves may appear and disappear unpredictably. Thus, an effective, automated search and selection of relevant services or agents is essential for human users and agents as well. We distinguish three general agent categories in the Cyberspace, service providers, service requester, and middle agents. Service providers provide some type of service, such as finding information, or performing some particular domain specific problem solving. Requester agents need provider agents to perform some service for them. Agents that help locate others are called middle agents[2]. Matchmaking is the process of finding an appropriate provider for a requester thr AI jeh02scaling Scaling Personalized Web Search Recent web search techniques augment traditional text matching with a global notion of “importance ” based on the linkage structure of the web, such as in Google’s PageRank algorithm. For more refined searches, this global notion of importance can be specialized to create personalized views of importance—for example, importance scores can be biased according to a user-specified set of initially-interesting pages. Computing and storing all possible personalized views in advance is impractical, as is computing personalized views at query time, since the computation of each view requires an iterative computation over the web graph. We present new graph-theoretical results, and a new technique based on these results, that encode personalized views as partial vectors. Partial vectors are shared across multiple personalized views, and their computation and storage costs scale well with the number of views. Our approach enables incremental computation, so that the construction of personalized views from partial vectors is practical at query time. We present efficient dynamic programming algorithms for computing partial vectors, an algorithm for constructing personalized views from partial vectors, and experimental results demonstrating the effectiveness and scalability of our techniques. 1 IR smola01regularized Regularized Principal Manifolds . Many settings of unsupervised learning can be viewed as quantization problems --- the minimization of the expected quantization error subject to some restrictions. This allows the use of tools such as regularization from the theory of (supervised) risk minimization for unsupervised settings. This setting turns out to be closely related to principal curves, the generative topographic map, and robust coding. We explore this connection in two ways: 1) we propose an algorithm for finding principal manifolds that can be regularized in a variety of ways. 2) We derive uniform convergence bounds and hence bounds on the learning rates of the algorithm. In particular, we give bounds on the covering numbers which allows us to obtain nearly optimal learning rates for certain types of regularization operators. Experimental results demonstrate the feasibility of the approach. Keywords: Regularization, Uniform Convergence, Kernels, Entropy Numbers, Principal Curves, Clustering, Generative Topograph... IR cardie99integrating Integrating Case-Based Learning and Cognitive Biases for Machine Learning of Natural Language This paper shows that psychological constraints on human information processing can be used effectively to guide feature set selection for case-based learning of linguistic knowledge. Given as input a baseline case representation for a natural language learning task, our algorithm selects the relevant cognitive biases for the task and then automatically modifies the representation in response to those biases by changing, deleting, and weighting features appropriately. We apply the cognitive bias approach to feature set selection to four natural language learning problems and show that performance of the casebased learning algorithm improves significantly when relevant cognitive biases are incorporated into the baseline instance representation. We argue that the cognitive bias approach offers new possibilities for case-based learning of natural language: it simplifies the process of instance representation design and, in theory, obviates the need for separate instance represent... ML 454876 A Study of Approaches to Hypertext Categorization Hypertext poses new research challenges for text classification. Hyperlinks, HTML tags, category labels distributed over linked documents, and meta data extracted from related Web sites all provide rich information for classifying hypertext documents. How to appropriately represent that information and automatically learn statistical patterns for solving hypertext classification problems is an open question. This paper seeks a principled approach to providing the answers. Specifically, we define five hypertext regularities which may (or may not) hold in a particular application domain, and whose presence (or absence) may significantly influence the optimal design of a classifier. Using three hypertext datasets and three well-known learning algorithms (Naive Bayes, Nearest Neighbor, and First Order Inductive Learner), we examine these regularities in different domains, and compare alternative ways to exploit them. Our results show that the identification of hypertext regularities in the data and the selection of appropriate representations for hypertext in particular domains are crucial, but seldom obvious, in real-world problems. We find that adding the words in the linked neighborhood to the page having those links (both inlinks and outlinks) were helpful for all our classifiers on one data set, but more harmful than helpful for two out of the three classifiers on the remaining datasets. We also observed that extracting meta data from related Web sites was extremely useful for improving classification accuracy in some of those domains. Finally, the relative performance of the classifiers being tested provided insights into their strengths and limitations for solving classification problems involving diverse and often noisy Web pages. IR 270469 Patterns of Search: Analyzing and Modeling Web Query Refinement . We discuss the construction of probabilistic models centering on temporal patterns of query refinement. Our analyses are derived from a large corpus of Web search queries extracted from server logs recorded by a popular Internet search service. We frame the modeling task in terms of pursuing an understanding of probabilistic relationships among temporal patterns of activity, informational goals, and classes of query refinement. We construct Bayesian networks that predict search behavior, with a focus on the progression of queries over time. We review a methodology for abstracting and tagging user queries. After presenting key statistics on query length, query frequency, and informational goals, we describe user models that capture the dynamics of query refinement. 1 Introduction The evolution of the World Wide Web has provided rich opportunities for gathering and analyzing anonymous log data generated by user interactions with network-based services. Web-based search engine... IR 335912 OBDD-based Universal Planning for Synchronized Agents in Non-Deterministic Domains Recently model checking representation and search techniques were shown to be efficiently applicable to planning, in particular to non-deterministic planning. Such planning approaches use Ordered Binary Decision Diagrams (obdds) to encode a planning domain as a non-deterministic finite automaton and then apply fast algorithms from model checking to search for a solution. obdds can effectively scale and can provide universal plans for complex planning domains. We are particularly interested in addressing the complexities arising in non-deterministic, multi-agent domains. In this article, we present umop, a new universal obdd-based planning framework for non-deterministic, multi-agent domains. We introduce a new planning domain description language, NADL, to specify non-deterministic, multi-agent domains. The language contributes the explicit definition of controllable agents and uncontrollable environment agents. We describe the syntax and semantics of NADL and show how to bu... ML moreno95alisis An'alisis Din'amico De Las Creencias El modelo de los mundos posibles y su sem'antica de Kripke asociada dan una sem'antica intuitiva para las l'ogicas dox'asticas, pero parecen llevar inevitablemente a modelizar agentes l'ogicamente omniscientes y razonadores perfectos. Estos problemas son evitables, si los mundos posibles dejan de considerarse como descripciones completas y consistentes del mundo real. Adoptando una definici'on sint'actica de mundos posibles, en este art'iculo se sugiere c'omo se pueden analizar las creencias de una forma puramente l'ogica, usando el m'etodo de los tableros anal'iticos (con ciertas modificaciones). Este an'alisis constituye un primer paso hacia la modelizaci'on de la investigaci'on racional. Palabras clave: l'ogicas de creencias y de conocimiento, omnisciencia l'ogica, mundos posibles, relaci'on de accesibilidad, tableros anal'iticos Temas: representaci'on del conocimiento, razonamiento 1 Introducci'on Las l'ogicas de creencias y de conocimiento ([HALP92], [FAGI95]) son herramienta... Agents roobaert00comparison Comparison of Learning Approaches to Appearance-based 3D Object Recognition with and without cluttered background We re-evaluate the application of Support Vector Machines (SVM) to appearance-based 3D object recognition, by comparing it to two other learning approaches: the system developed at Columbia University ("Columbia") and a simple image matching system using a nearest neighbor classifier ("NNC"). In a first set of experiments, we compare correct recognition rates of the segmented 3D object images of the COIL database. We show that the performance of the simple "NNC" system compares to the more elaborated "Columbia" and "SVM" systems. Only when the experimental setting is more demanding, i.e. when we reduce the number of views during the training phase, some difference in performance can be observed. In a second set of experiments, we consider the more realistic task of 3D object recognition with cluttered background. Also in this case, we obtain that the performance of the three systems are comparable. Only with the recently proposed black/white background training scheme ("BW") applied t... ML prasad96distributed Distributed Case-Based Learning Multi-agent systems exploiting case based reasoning techniques have to deal with the problem of retrieving episodes that are themselves distributed across a set of agents. From a Gestalt perspective, a good overall case may not be the one derived from the summation of best subcases. In this paper we deal with issues involved in learning and exploiting the learned knowledge in multiagent case-based systems. Introduction Case Based Reasoning (CBR) has been attracting much attention recently as a paradigm with a wide variety of applications [Kolodner 93]. In this paper, we discuss issues pertaining to cooperative retrieval and composition of a case in which subcases are distributed across different agents in a multi-agent system. A multi-agent system comprises a group of intelligent agents working towards a set of common global goals or separate individual goals that may interact. In such a system, each of the agents may not be individually capable of achieving the global goal and/or ... ML sun99optional Optional Locking Integrated with Operational Transformation in Distributed Real-Time Group Editors Locking is a standard technique in traditional distributed computing and database systems to ensure data integrity by prohibiting concurrent conflicting updates on shared data objects. Operational transformation is an innovative technique invented by groupware research for consistency maintenance in real-time group editors. In this paper, we will examine and explore the complementary roles of locking and operational transformation in consistency maintenance. A novel optional locking scheme is proposed and integrated with operation transformation to maintain both generic and context-specific consistency in a distributed, interactive, and collaborative environment. The integrated optional locking and operational transformation technique is fully distributed, highly responsive, non-blocking, and capable of avoiding locking overhead in the most common case of collaborative editing. Keywords: Locking, operational transformation, consistency maintenance, group editors, groupware, distribute... DB pfister98room Rooms, Protocols, and Nets: Metaphors for Computer Supported Cooperative Learning of Distributed Groups : We discuss an integrative design for computer supported cooperative learning (CSCL) environments. Three common problems of CSCL are addressed: How to achieve social orientation and group awareness, how to coordinate goal-directed interaction, and how to construct a shared knowledge base. With respect to each problem, we propose a guiding metaphor which links theoretical, technical, and usability requirements. If appropriately implemented, each metaphor resolves one problem: Virtual rooms support social orientation, learning protocols guide interactions aimed at knowledge acquisition, and learning nets represent socially shared knowledge. Theoretically, the metaphor of virtual rooms originates in work on virtual spaces in human computer interaction, learning protocols are related to speech act theory, and learning nets are based on models of knowledge representation. A prototype system implementing the virtual room metaphor is presented. We argue that by further integrating these thre... HCI chen00algebraic An Algebraic Compression Framework for Query Results Decision-support applications in emerging environments require that SQL query results or intermediate results be shipped to clients for further analysis and presentation. These clients may use low bandwidth connections or have severe storage restrictions. Consequently, there is a need to compress the results of a query for efficient transfer and client-side access. This paper explores a variety of techniques that address this issue. Instead of using a fixed method, we choose a combination of compression methods that use statistical and semantic information of the query results to enhance the effect of compression. To represent such a combination, we present a framework of "compression plans" formed by composing primitive compression operators. We also present optimization algorithms that enumerate valid compression plans and choose an optimal plan. Our experiments show that our techniques achieve significant performance improvement over standard compression tools like WinZip. 1. Intro... DB ormoneit01learning Learning and Tracking Cyclic Human Motion We present methods for learning and tracking human motion in video. We estimate a statistical model of typical activities from a large set of 3D periodic human motion data by segmenting these data automatically into "cycles". Then the mean and the principal components of the cycles are computed using a new algorithm that accounts for missing information and enforces smooth transitions between cycles. The learned temporal model provides a prior probability distribution over human motions that can be used in a Bayesian framework for tracking human subjects in complex monocular video sequences and recovering their 3D motion. HCI combs99does Does Zooming Improve Image Browsing? We describe an image retrieval system we built based on a Zoomable User Interface (ZUI). We also discuss the design, results and analysis of a controlled experiment we performed on the browsing aspects of the system. The experiment resulted in a statistically significant difference in the interaction between number of images (25, 75, 225) and style of browser (2D, ZUI, 3D). The 2D and ZUI browser systems performed equally, and both performed better than the 3D systems. The image browsers tested during the experiment include Cerious Software's Thumbs Plus, TriVista Technology's Simple LandScape and Photo GoRound, and our Zoomable Image Browser based on Pad++. Keywords Evaluation, controlled experiment, image browsers, retrieval systems, real-time computer graphics, Zoomable User Interfaces (ZUIs), multiscale interfaces, Pad++. INTRODUCTION In the past two decades, with the emergence of faster computers, the declining cost of memory, the popularity of digital cameras, online archives... HCI veloso94learning Learning Strategy Knowledge Incrementally Modern industrial processes require advanced computer tools that should adapt to the user requirements and to the tasks being solved. Strategy learning consists of automating the acquisition of patterns of actions used while solving particular tasks. Current intelligent strategy learning systems acquire operational knowledge to improve the efficiency of a particular problem solver. However, these strategy learning tools should also provide a way of achieving low-cost solutions according to user-specific criteria. In this paper, we present a learning system, hamlet, which is integrated in a planning architecture, prodigy, and acquires control knowledge to guide prodigy to efficiently produce cost-effective plans. hamlet learns from planning episodes, by explaining why the correct decisions were made, and later refines the learned strategy knowledge to make it incrementally correct with experience. ML panzarasa99modeling Modeling Sociality In The BDI Framework . We present a conceptual model for how the social nature of agents impacts upon their individual mental states. Roles and social relationships provide an abstraction upon which we develop the notion of social mental shaping . 1 Introduction Belief-Desire-Intention (BDI) architectures for deliberative agents are based on the physical symbol system assumption that agents maintain and reason about internal representations of their world [2]. However, while such architectures conceptualise individual intentionality and behaviour, they say nothing about the social aspects of agents being situated in a multi-agent system. The main reason for this limitation is that mental attitudes are taken to be internal to a particular agent (or team) and are modeled as a relation between the agent (or a team) and a proposition. The purpose of this paper is, therefore, to extend BDI models in order to investigate the problem of how the social nature of agents can impact upon their individual mental ... Agents shaw98clause Clause Aggregation Using Linguistic Knowledge By combining multiple clauses into one single sentence, a text generation system can express the same amount of information in fewer words and at the same time, produce a great variety of complex constructions. In this paper, we describe hypotactic and paratactic operators for generating complex sentences from clause-sized semantic representations. These two types of operators are portable and reusable because they are based on general resources such as the lexicon and the grammar. 1 Introduction An expression is more concise than another expression if it conveys the same amount of information in fewer words. Complex sentences generated by combining clauses are more concise than corresponding simple sentences because multiple references to the recurring entities are removed. For example, clauses like "Jones is a patient" and "Jones has hypertension" can be combined into a more concise sentence "Jones is a hypertensive patient." To illustrate the common occurrence of such repeated enti... DB abraham00hybrid Hybrid Heuristics for Optimal Design of Artificial Neural Networks Designing the architecture and correct parameters for the learning algorithm is a tedious task for modeling an optimal Artificial Neural Network (ANN), which is smaller, faster and with a better generalization performance. In this paper we explain how a hybrid algorithm integrating Genetic algorithm (GA), Simulated Annealing (SA) and other heuristic procedures can be applied for the optimal design of an ANN. This paper is more concerned with the understanding of current theoretical developments of Evolutionary Artificial Neural Networks (EANNs) using GAs and how the proposed hybrid heuristic procedures can be combined to produce an optimal ANN. The proposed meta-heuristic can be regarded as a general framework for adaptive systems, that is, systems that can change their connection weights, architectures and learning rules according to different environments without human intervention. AI cummins99automatic Automatic Discrimination Among Languages Based on Prosody Alone The development of methods for the automatic identification of languages is motivated both by speech-based applications intended for use in a multi-lingual environment, and by theoretical questions of cross-linguistic variation and similarity. We evaluate the potential utility of two prosodic variables, F 0 and amplitude envelope modulation, in a pairwise language discrimination task. Discrimination is done using a novel neural network which can successfully attend to temporal information at a range of timescales. Both variables are found to be useful in discriminating among languages, and confusion patterns, in general, reflect traditional intonational and rhythmic language classes. The methods employed allow empirical determination of prosodic similarity across languages. Die Entwicklung von Methoden zur automatischen Sprachidentifikation wird motiviert sowohl durch sprach-basierte Anwendungen, die zum Einsatz in einer mehrsprachigen Umgebung bestimmt sind, als auch durch theoretisch... ML 57340 Class Representation and Image Retrieval with Non-Metric Distances One of the key problems in appearance-based vision is understanding how to use a set of labeled images to classify new images. Classification systems that can model human performance, or that use robust image matching methods, often make use of similarity judgments that are non-metric; but when the triangle inequality is not obeyed, most existing pattern recognition techniques are not applicable. We note that exemplar-based (or nearest-neighbor) methods can be applied naturally when using a wide class of non-metric similarity functions. The key issue, however, is to find methods for choosing good representatives of a class that accurately characterize it. We show that existing condensing techniques for finding class representatives are ill-suited to deal with non-metric dataspaces. We then focus on developing techniques for solving this problem, emphasizing two points: First, we show that the distance between two images is not a good measure of how well one image can represent another in non-metric spaces. Instead, we use the vector correlation between the distances from each image to other previously seen images. Second, we show that in non-metric spaces, boundary points are less significant for capturing the structure of a class ML 451065 Ontology-Related Services in Agent-Based Distributed Information Infrastructures Ontologies are an emerging paradigm to support declarativity, interoperability, and intelligent services in many areas, such as Agent--based Computation, Distributed Information Systems, and Expert Systems. In the context of designing a scalable, agent-based middleware for the realization of distributed Organizational Memories (OM), we examine the question what ontology--related services must be provided as middleware components. To this end, we discuss three basic dimensions of information that have fundamental impact on the usefulness of ontologies for OMs, namely formality, stability, and sharing scope of information. A short discussion of techniques which are suited to find a balance in each of these dimensions leads to a characterization of roles of ontology--related actors in the OM scenario. We describe the several roles with respect to their goals, knowledge, competencies, rights, and obligations. These actor classes and the related competencies are candidates to define agent types, speech acts, and standard services in the envisioned OM middleware. 1. AI emmerich99incremental Incremental Code Mobility with XML We demonstrate how XML and related technologies can be used for code mobility at any granularity, thus overcoming the restrictions of existing approaches. By not fixing a particular granularity for mobile code, we enable complete programs as well as individual lines of code to be sent across the network. We define the concept of incremental code mobility as the ability to migrate and add, remove, or replace code fragments (i.e., increments) in a remote program. The combination of fine-grained and incremental mobility achieves a previously unavailable degree of flexibility. We examine the application of incremental and fine-grained code mobility to a variety of domains, including user interface management, application management on mobile thin clients, for example PDAs, and management of distributed documents. Keywords Incremental Code Mobility, XML Technologies 1 INTRODUCTION The increasing popularity of Java and the spread of Webbased technologies are contributing to a growing int... HCI 301461 Organisational Abstractions for the Analysis and Design of Multi-Agent Systems Abstract. The architecture of a multi-agent system can naturally be viewed as a computational organisation. For this reason, we believe organisational abstractions should play a central role in the analysis and design of such systems. To this end, the concepts of agent roles and role models are increasingly being used to specify and design multi-agent systems. However, this is not the full picture. In this paper we introduce three additional organisational concepts — organisational rules, organisational structures, and organisational patterns — that we believe are necessary for the complete specification of computational organisations. We view the introduction of these concepts as a step towards a comprehensive methodology for agent-oriented systems. 1 Agents hannebauer99dynamic Dynamic Reconfiguration in Collaborative Problem Solving In this article we will describe our research efforts in coping with a trade-off that can be often found in the control and optimization of todays business processes. Though centralized control may achieve nearto -optimum results in optimizing the system behavior, there are usually social, technical and security restrictions on applying centralized control. Distributed control on the other hand may cope with these restrictions but also entails sub-optimality and communicational overhead. Our concept of composable agents tries to allow a dynamic and fluent transition between globalization and localization in business process control by adapting to the current real-world system structure. We are currently evaluating this concept in the framework of patient flow control at Charit'e Berlin. Introduction Research in Distributed Artificial Intelligence (DAI, (Bond & Gasser 1988)) has been traditionally divided into Distributed Problem Solving (DPS) and Multi Agent Systems (MAS). However, r... Agents 296568 Combining Multiple Learning Strategies for Effective Cross Validation Parameter tuning through cross-validation becomes very difficult when the validation set contains no or only a few examples of the classes in the evaluation set. We address this open challenge by using a combination of classifiers with different performance characteristics to effectively reduce the performance variance on average of the overall system across all classes, including those not seen before. This approach allows us to tune the combination system on available but less-representative validation data and obtain smaller performance degradation of this system on the evaluation data than using a single-method classifier alone. We tested this approach by applying k-Nearest Neighbor, Rocchio and Language Modeling classifiers and their combination to the event tracking problem in the Topic Detection and Tracking (TDT) domain, where new classes (events) are created constantly over time, and representative validation sets for new classes are often difficult to ob... IR alvarez00web Web Metasearch as Belief Aggregation Web metasearch requires a mechanism for combining rank-ordered lists of ratings returned by multiple search engines in response to a given user query. We view this as being analogous to the need for combining degrees of belief in probabilistic and uncertain reasoning in artificial intelligence. This paper describes a practical method for performing web metasearch based on a novel transformationbased theory of belief aggregation. The consensus ratings produced by this method take into account the item ratings/rankings output by individual search engines as well as the user's preferences. Copyright c fl 2000, American Association for Artificial Intelligence (www.aaai.org). All rights reserved. Introduction Web search engines (WSE) use tools ranging from simple text-based search to more sophisticated methods that attempt to understand the intended meanings of both queries and data items. There has been much work in this area in recent years. The link structure of the web has... IR vandendoel01foleyautomatic FOLEYAUTOMATIC: Physically-based Sound Effects for Interactive Simulation and Animation Animations for which sound effects were automatically added by our system, demonstrated in the accompanying video. (a) A real wok in which a pebble is thrown; the pebble rattles around the wok and comes to rest after wobbling. (b) A simulation of a pebble thrown in wok, with all sound effects automatically generated. (c) A ball rolling back and forth on a ribbed surface. (d) Interaction with a sonified object. We describe algorithms for real-time synthesis of realistic sound effects for interactive simulations (e.g., games) and animation. These sound effects are produced automatically, from 3D models using dynamic simulation and user interaction. We develop algorithms that are efficient, physicallybased, and can be controlled by users in natural ways. We develop effective techniques for producing high quality continuous contact sounds from dynamic simulations running at video rates which are slow relative to audio synthesis. We accomplish this using modal models driven by contact forces modeled at audio rates, which are much higher than the graphics frame rate. The contact forces can be computed from simulations or can be custom designed. We demonstrate the effectiveness with complex realistic simulations. HCI maneewongvatana99analysis Analysis of Approximate Nearest Neighbor Searching with Clustered Point Sets this paper we study the performance of two other splitting methods, and compare them against the kd-tree splitting method. The first, called slidingmidpoint, is a splitting method that was introduced by Mount and Arya in the ANN library for approximate nearest neighbor searching [30]. This method was introduced into the library in order to better handle highly clustered data sets. We know of no analysis (empirical or theoretical) of this method. This method was designed as a simple technique for addressing one of the most serious flaws in the standard kd-tree splitting method. The flaw is that when the data points are highly clustered in low dimensional subspaces, then the standard kd-tree splitting method may produce highly elongated cells, and these can lead to slow query times. This splitting method starts with a simple midpoint split of the longest side of the cell, but if this split results in either subcell containing no data points, it translates (or "slides") the splitting plane in the direction of the points until hitting the first data point. In Section 3.1 we describe this splitting method and analyze some of its properties. The second splitting method, called minimum-ambiguity, is a query-based technique. The tree is given not only the data points, but also a collection of sample query points, called the training points. The algorithm applies a greedy heuristic to build the tree in an attempt to minimize the expected query time on the training points. We model query processing as the problem of eliminating data points from consideration as the possible candidates for the nearest neighbor. Given a collection of query points, we can model any stage of the nearest neighbor algorithm as a bipartite graph, called the candidate graph, whose vertices correspond t... ML gelati01agents Agents Supporting Information Integration: The Miks Framework During past years we have developed the MOMIS (Mediator envirOnment for Multiple Information Sources) system for the integration of data from structured and semi-structured data sources. In this paper we propose a new system, MIKS (Mediator agent for Integration of Knowledge Sources), which enriches the MOMIS architecture exploiting the intelligent and mobile agent features. 1. Motivation The web explosion, both at Internet and intranet level, has transformed the electronic information system from single isolated node to an entry point into a worldwide network of information exchange and business transactions. One of the main challenges for the designers of the e-commerce infrastructures is the information sharing, retrieving data located in different sources thus obtaining an integrated view to overcome any contradiction or redundancy. During past years we have developed the MOMIS (Mediator envirOnment for Multiple Information Sources) system for the integration of data from struc... Agents 28981 Relational Learning Techniques for Natural Language Information Extraction The recent growth of online information available in the form of natural language documents creates a greater need for computing systems with the ability to process those documents to simplify access to the information. One type of processing appropriate for many tasks is information extraction, a type of text skimming that retrieves specific types of information from text. Although information extraction systems have existed for two decades, these systems have generally been built by hand and contain domain specific information, making them difficult to port to other domains. A few researchers have begun to apply machine learning to information extraction tasks, but most of this work has involved applying learning to pieces of a much larger system. This paper presents a novel rule representation specific to natural language and a learning system, Rapier, which learns information extraction rules. Rapier takes pairs of documents and filled templates indicating the information to be ext... DB brabrand99ltbigwiggt <bigwig> -- A language for developing interactive Web services <bigwig> is a high-level programming language and a compiler for developing interactive Web services. The overall goal of the language design is to remove many of the obstacles that face current developers of Web services in order to lower cost while increasing functionality and reliability. The compiler translates programs into a conglomerate of lower-level standard technologies such as CGI-scripts, HTML, JavaScript, and HTTP Authentication. This paper describes the major facets of the language design and the techniques used in their implementation, and compares the design with alternative Web service technologies. DB zhou00implementation Implementation of a Linear Tabling Mechanism Delaying-based tabling mechanisms, such as the one adopted in XSB, are nonlinear in the sense that the computation state of delayed calls has to be preserved. DB tice00key Key Instructions: Solving the Code Location Problem for Optimized Code There are many difficulties to be overcome in the process of designing and implementing a debugger for optimized code. One of the first problems facing the designer of such a debugger is determining how to accurately map between locations in the source program and locations in the corresponding optimized binary. The solution to this problem is critical for many aspects of debugger design, from setting breakpoints, to implementing single-stepping, to reporting error locations. Previous approaches to debugging optimized code have presented many different techniques for solving this location mapping problem (commonly known as the code location problem). These techniques are often very complex and sometimes incomplete. Identifying key instructions allows for a simple yet formal way of mapping between locations in the source program and the optimized target program. In this paper we present the concept of key instructions. We give a formal definition of key instructions and present algorit... HCI forlizzi99data A Data Model and Data Structures for Moving Objects Databases We consider spatio-temporal databases supporting spatial objects with continuously changing position and extent, termed moving objects databases. We formally define a data model for such databases that includes complex evolving spatial structures such as line networks or multi-component regions with holes. The data model is given as a collection of data types and operations which can be plugged as attribute types into any DBMS data model (e.g. relational, or object-oriented) to obtain a complete model and query language. A particular novel concept is the sliced representation which represents a temporal development as a set of units, where unit types for spatial and other data types represent certain "simple" functions of time. We also show how the model can be mapped into concrete physical data structures in a DBMS environment. 1 Introduction A wide and increasing range of database applications has to deal with spatial objects whose position and/or extent changes over time... DB himberg01time Time Series Segmentation for Context Recognition in Mobile Devices Recognizing the context of use is important in making mobile devices as simple to use as possible. Finding out what the user's situation is can help the device and underlying service in providing an adaptive and personalized user interface. The device can infer parts of the context of the user from sensor data: the mobile device can include sensors for acceleration, noise level, luminosity, humidity, etc. In this paper we consider context recognition by unsupervised segmentation of time series produced by sensors. Dynamic programming can be used to find segments that minimize the intra-segment variances. While this method produces optimal solutions, it is too slow for long sequences of data. We present and analyze randomized variations of the algorithm. One of them, Global Iterative Replacement or GIR, gives approximately optimal results in a fraction of the time required by dynamic programming. We demonstrate the use of time series segmentation in context recognition for mobile phone applications. 1 HCI ng01stable Stable Algorithms for Link Analysis The Kleinberg HITS and the Google PageRank algorithms are eigenvector methods for identifying "authoritative" or "influential" articles, given hyperlink or citation information. That such algorithms should give reliable or consistent answers is surely a desideratum, and in [10], we analyzed when they can be expected to give stable rankings under small perturbations to the linkage patterns. In this paper, we extend the analysis and show how it gives insight into ways of designing stable link analysis methods. This in turn motivates two new algorithms, whose performance we study empirically using citation data and web hyperlink data. 1. IR yeates01memory Memory Hierarchies as a Metaphor for Academic Library Collections Research libraries and their collections are a cornerstone of the academic tradition, representing 2000 years of development of the Western Civilization; they make written history widely accessible at low cost. Computer memories are a range of physical devices used for storing digital information that have undergone much formal study over 40 years and are well understood. This paper draws parallels between the organisation of research collections and computer memories, in particular examining their hierarchical structure, and examines the implication for digital libraries. IR 231484 Autonomous Evolution of Gaits with the Sony Quadruped Robot A trend in robotics is towards legged robots. One of the issues with legged robots is the development of gaits. Typically gaits are developed manually. In this paper we report our results of autonomous evolution of dynamic gaits for the Sony Quadruped Robot. Fitness is determined using the robot's digital camera and infrared sensors. Using this system we evolve faster dynamic gaits than previously manually developed. 1 INTRODUCTION In this paper we present an implementation of an autonomous evolutionary algorithm (EA) for developing locomotion gaits. All processing is handled by the robot's onboard computer and individuals are evaluated using the robot's sensors. Our implementation successfully evolves trot and pace gaits for our robot for which the pace gait significantly outperforms previous hand-developed gaits. In addition to achieving our desired goal of automatically developing gaits these results show that EAs can be used on real robots to evolve non-trivial behaviors. A method... ML 462179 Infinite-Horizon Policy-Gradient Estimation Gradient-based approaches to direct policy search in reinforcement learning have received much recent attention as a means to solve problems of partial observability and to avoid some of the problems associated with policy degradation in value-function methods. In this paper we introduce GPOMDP, a simulation-based algorithm for generating a biased estimate of the gradient of the average reward in Partially Observable Markov Decision Processes (POMDPs) controlled by parameterized stochastic policies. A similar algorithm was proposed by Kimura, Yamamura, and Kobayashi (1995). The algorithm's chief advantages are that it requires storage of only twice the number of policy parameters, uses one free parameter 2 [0; 1) (which has a natural interpretation in terms of bias-variance trade-off), and requires no knowledge of the underlying state. We prove convergence of GPOMDP, and show how the correct choice of the parameter is related to the mixing time of the controlled POMDP. We briefly describe extensions of GPOMDP to controlled Markov chains, continuous state, observation and control spaces, multiple-agents, higher-order derivatives, and a version for training stochastic policies with internal states. In a companion paper (Baxter, Bartlett, & Weaver, 2001) we show how the gradient estimates generated by GPOMDP can be used in both a traditional stochastic gradient algorithm and a conjugate-gradient procedure to find local optima of the average reward. ML jr99identifying Identifying and Handling Structural Incompleteness for Validation of Probabilistic Knowledge-Bases The PESKI (Probabilities, Expert Systems, Knowledge, and Inference) system attempts to address some of the problems in expert system design through the use of the Bayesian Knowledge Base (BKB) representation. AI broekstra02enabling Enabling knowledge representation on the Web by extending RDF Schema Recently, there has been a wide interest in using ontologies on the Web. As a basis for this, RDF Schema (RDFS) provides means to define vocabulary, structure and constraints for expressing metadata about Web resources. However, formal semantics are not provided, and the expressivity of it is not enough for full-fledged ontological modeling and reasoning. In this paper, we will show how RDFS can be extended in such a way that a full knowledge representation (KR) language can be expressed in it, thus enriching it with the required additional expressivity and the semantics of this language. We do this by describing the ontology language OIL as an extension of RDFS. An important advantage of our approach is a maximal backward compatability with RDFS: any meta-data in OIL format can still be partially interpreted by any RDFS-only-processor. The OIL extension of RDFS has been carefully engineered so that such a partial interpretation of OIL meta-data is still correct under the intended semantics of RDFS: simply ignoring the OIL specific portions of an OIL document yields a correct RDF(S) document whose intended RDFS semantics is precisely a subset of the semantics of the full OIL statements. In this way, our approach ensures maximal sharing of meta-data on the Web: even partial interpretation of meta-data by less semantically aware processors will yield a correct partial interpretation of the meta-data. We conclude that our method of extending is equally applicable to other KR formalisms. 1 1 IR itti99comparison A Comparison of Feature Combination Strategies for Saliency-Based Visual Attention Systems Bottom-up or saliency-based visual attention allows primates to detect non-specific conspicuous targets in cluttered scenes. A classical metaphor, derived from electrophysiological and psychophysical studies, describes attention as a rapidly shiftable "spotlight". The model described here reproduces the attentional scanpaths of this spotlight: Simple multi-scale "feature maps" detect local spatial discontinuities in intensity, color, orientation or optical flow, and are combined into a unique "master" or "saliency" map. The saliency map is sequentially scanned, in order of decreasing saliency, by the focus of attention. We study the problem of combining feature maps, from different visual modalities and with unrelated dynamic ranges (such as color and motion), into a unique saliency map. Four combination strategies are compared using three databases of natural color images: (1) Simple normalized summation, (2) linear combination with learned weights, (3) global non-linear normalization... ML lind01issues Issues in Agent-Oriented Software Engineering In this paper, I will discuss the conceptual foundation of agent-oriented software development by relating the fundamental elements of the agent-oriented view to those of other, well established programming paradigms, especially the object-oriented approach. Furthermore, I will motivate the concept of autonomy as the basic property of the agent-oriented school and discuss the development history of programming paradigms that lead to this perspective on software systems. The paper will be concluded by an outlook on how the new paradigm can change the way we think about software systems. DB 233508 Developments in Spatio-Temporal Query Languages In contrast to the field view of spatial data that basically views spatial data as a mapping from points into some features, the object view clusters points by features and their values into spatial objects of type point, line, or region. When embedding these objects into a data model, such as the relational model, an additional clustering according to conceptually identified objects takes place. For example, we could define a relation City(name: string,center: point,area: region) that combines different features for cities in one relation. An important aspect of this kind of modeling is that clustering happens on two different levels: (i) points are grouped into spatial objects like regions and (ii) different attributes/features are grouped into a perceived object. When talking about data modeling there is no reason why this grouping should be limited to two levels. For example, we can consider storing regions of different population densities for each city in an attribute density: num → region. Although then the relation is not in first normal form anymore, we can “recover” the first normal form by encapsulating the function num → region in an abstract data type. The important aspect is that all the required operations on such a type as well as on regions and other complex types can be defined to a large degree independently from the data model. 1 The most important point about the preceding discussion is the way in which complex types can be easily DB casillas00methodology A Methodology to Improve Ad Hoc Data-Driven Linguistic Rule Learning Methods by Inducing Cooperation Among Rules Within the Linguistic Modeling eld |one of the most important applications of Fuzzy Rule-Based Systems|, a family of ecient and simple methods guided by covering criteria of the data in the example set, called \ad hoc data-driven methods", has been proposed in the literature in the last few years. Their high performance, in addition to their quickness and easy understanding, have make them very suitable for learning tasks. In this paper we are going to perform a double task analyzing these kinds of learning methods and introducing a new methodology to signicantly improve their accuracy keeping their descriptive power unalterable. On the one hand, a taxonomy of ad hoc data-driven learning methods based on the way in which the available data is used to guide the learning will be made. In this sense, we will distinguish between two approaches: the example-based and the fuzzy-grid-based one. Whilst in the former each rule is obtained from a specic example, in the latter the e... ML ulrich00appearancebased Appearance-Based Obstacle Detection with Monocular Color Vision This paper presents a new vision-based obstacle detection method for mobile robots. Each individual image pixel is classified as belonging either to an obstacle or the ground based on its color appearance. The method uses a single passive color camera, performs in real-time, and provides a binary obstacle image at high resolution. The system is easily trained by simply driving the robot through its environment. In the adaptive mode, the system keeps learning the appearance of the ground during operation. The system has been tested successfully in a variety of environments, indoors as well as outdoors. 1. Introduction Obstacle detection is an important task for many mobile robot applications. Most mobile robots rely on range data for obstacle detection. Popular sensors for range-based obstacle detection systems include ultrasonic sensors, laser rangefinders, radar, stereo vision, optical flow, and depth from focus. Because these sensors measure the distances from obstacles t... ML steiner98omsjava OMS/Java: Model Extensibility of OODBMS for Advanced Application Domains . We showhow model extensibility of object-oriented data management systems can be achieved through the combination of a highlevel core object data model and an architecture designed with model extensibility in mind. The resulting system, OMS#Java, is both a general data management system and a framework for the developmentof advanced database application systems. All aspects of the core model # constructs, query language and constraints # can easily be generalised to support, for example, the management of temporal, spatial and versioned data. Speci#cally,we showhow the framework was used to extend the core system to a temporal object-oriented database management system. 1 Introduction Extensibility has often been considered a purely architectural issue in database management systems #DBMS#. In the 1980s, there was an increase in the various forms of DBMS that appeared --- many of whichwere tailored to speci#c application domains such as Geographical Information Systems or ... DB kueng01holistic A Holistic Process Performance Analysis through a Performance Data Warehouse This paper describes how a performance data warehouse can be used to facilitate business process improvement that is based on holistic performance measurement. The feasibility study shows how management and analysis of performance data can be facilitated by a data warehouse approach. It is argued that many of the shortcomings of traditional measurement systems can be overcome with this performance data warehouse approach. DB stone98layered A Layered Approach to Learning Client Behaviors in the RoboCup Soccer Server In the past few years, Multiagent Systems (MAS) has emerged as an active subfield of Artificial Intelligence (AI). Because of the inherent complexity of MAS, there is much interest in using Machine Learning (ML) techniques to help build multiagent systems. Robotic soccer is a particularly good domain for studying MAS and Multiagent Learning. Our approach to using ML as a tool for building Soccer Server clients involves layering increasingly complex learned behaviors. In this article, we describe two levels of learned behaviors. First, the clients learn a low-level individual skill that allows them to control the ball effectively. Then, using this learned skill, they learn a higher-level skill that involves multiple players. For both skills, we describe the learning method in detail and report on our extensive empirical testing. We also verify empirically that the learned skills are applicable to game situations. 1 Introduction In the past few years, Multiagent Systems (MAS) has emerge... AI 472783 A Meta-search Method Reinforced by Cluster Descriptors A meta-search engine acts as an agent for the participant search engines. It receives queries from users and redirects them to one or more of the participant search engines for processing. A meta-search engine incorporating many participant search engines is better than a single global search engine in terms of the number of pages indexed and the freshness of the indexes. The meta-search engine stores descriptive data (i.e., descriptors) about the index maintained by each participant search engine so that it can estimate the relevance of each search engine when a query is received. The ability for the meta-search engine to select the most relevant search engines determines the quality of the final result. To facilitate the selection process, the document space covered by each search engine must be described not only concisely but also precisely. Existing methods tend to focus on the conciseness of the descriptors by keeping a descriptor for a search engine 's entire index. This paper proposes to cluster a search engine's document space into clusters and keep a descriptor for each cluster. We show that cluster descriptors can provide a finer and more accurate representation of the document space, and hence enable the meta-search engine to improve the selection of relevant search engines. Two cluster-based search engine selection scenarios (i.e., independent and high-correlation) are discussed in this paper. Experiments verify that the cluster-based search engine selection can effectively identify the most relevant search engines and improve the quality of the search results consequently. 1 IR skow02security A Security Architecture for Application Session Handoff Ubiquitous computing across a variety of wired and wireless connections still lacks an effective security architecture. In our research work, we address the specific issue of designing and building a security architecture for Application Session Handoff, a functionality which we envision will be a key component enabling ubiquitous computing. Our architecture incorporates a number of proven approaches into the new context of ubiquitous computing. We employ the Bell-LaPadula and capability models to realise access control and adopt Public Key Infrastructure (PKI)-based approaches to provide efficient and authenticated end-to-end security. To demonstrate the effectiveness of our design, we implemented an application enabled with this security architecture and showed that it incurred low latency. HCI karlgren98iterative Iterative Information Retrieval Using Fast Clustering and Usage-Specific Genres This paper describes how collection specific empirically defined stylistics based genre prediction can be brought together together with rapid topical clustering to build an interactive information retrieval interface with multi-dimensional presentation of search results. The prototype presented addresses two specific problems of information retrieval: how to enrich the information seeking dialog by encouraging and supporting iterative refinement of queries, and how to enrich the document representation past the shallow semantics allowed by term frequencies. Searching For More Than Words Today's tools for searching information in a document database are based on term occurrence in texts. The searcher enters a number of terms and a number of documents where those terms or closely related terms appear comparatively frequently are retrieved and presented by the system in list form. This method works well up to a point. It is intuitively understandable, and for competent users and well e... IR 488116 Agent-mediated Electronic Commerce: Scientific and Technological roadmap. this report is that a big part of Internet users have already sampled buying over the web (f.i. 40% in the UK), and a significat part qualify themselves as regular shoppers (f.i. 10% in the UK). Again, important differences between countries may be detected with respect to the expenses produced. For instance, Finland spent 20 times more that Spain on a per capita basis. The forecasts for European buying goods and services for the year 2002 suggest that the current 5.2 million shoppers will increase until 28.8 millions, and the European revenues from the current EUR3,032 million to EUR57,210 million. Finally, a significant increase in the number of European executives that believe in the future of electronic commerce has been observed (33% in 1999 up from 23% in 1998) Agents abonyi01modified Modified Gath-Geva Fuzzy Clustering for Identification of Takagi-Sugeno Fuzzy Models The construction of interpretable Takagi--Sugeno (TS) fuzzy models by means of clustering is addressed. First, it is shown how the antecedent fuzzy sets and the corresponding consequent parameters of the TS model can be derived from clusters obtained by the Gath--Geva algorithm. To preserve the partitioning of the antecedent space, linearly transformed input variables can be used in the model. This may, however, complicate the interpretation of the rules. To form an easily interpretable model that does not use the transformed input variables, a new clustering algorithm is proposed, based on the Expectation Maximization (EM) identification of Gaussian mixture models. This new technique is applied to two well-known benchmark problems: the MPG (miles per gallon) prediction and a simulated second-order nonlinear process. The obtained results are compared with results from the literature. ML blockeel00executing Executing Query Packs in ILP Inductive logic programming systems usually send large numbers of queries to a database. The lattice structure from which these queries are typically selected causes many of these queries to be highly similar. As a consequence, independent execution of all queries may involve a lot of redundant computation. We propose a mechanism for executing a hierarchically structured set of queries (a "query pack") through which a lot of redundancy in the computation is removed. We have incorporated our query pack execution mechanism in the ILP systems Tilde and Warmr by implementing a new Prolog engine ilProlog which provides support for pack execution at a lower level. Experimental results demonstrate significant efficiency gains. Our query pack execution mechanism is very general in nature and could be incorporated in most other ILP systems, with similar efficiency improvements to be expected. DB dixon00resolutionbased Resolution-Based Proof for Multi-Modal Temporal Logics of Knowledge Temporal logics of knowledge are useful in order to specify complex systems in which agents are both dynamic and have information about their surroundings. We present a resolution method for propositional temporal logic combined with multi-modal S5 and illustrate its use on examples. This paper corrects a previous proposal for resolution in multi-modal temporal logics of knowledge. Keywords: temporal and modal logics, non-classical resolution, theorem-proving 1 Introduction Combinations of logics have been useful for specifying and reasoning about complex situations, for example multi-agent systems [21, 24], accident analysis [15], and security protocols [18]. For example, logics to formalise multi-agent systems often incorporate a dynamic component representing change of over time; an informational component to capture the agent's knowledge or beliefs; and a motivational component for notions such as goals, wishes, desires or intentions. Often temporal or dynamic logic is used for... Agents castillo99gpropii G-Prop-II: Global Optimization of Multilayer Perceptrons using GAs A general problem in model selection is to obtain the right parameters that make a model fit observed data. For a Multilayer Perceptron (MLP) trained with Backpropagation (BP), this means finding appropiate layer size and initial weights. This paper proposes a method (G-Prop, genetic backpropagation) that attempts to solve that problem by combining a genetic algorithm (GA) and BP to train MLPs with a single hidden layer. The GA selects the initial weights and changes the number of neurons in the hidden layer through the application of specific genetic operators. G-Prop combines the advantages of the global search performed by the GA over the MLP parameter space and the local search of the BP algorithm. The application of the G-Prop algorithm to several real-world and benchmark problems shows that MLPs evolved using G-Prop are smaller and achieve a higher level of generalization than other perceptron training algorithms, such as QuickPropagation or RPROP, and other evolutive algorithms, s... ML federici95advances Advances in Analogy-Based Learning: False Friends and Exceptional Items in Pronunciation By Paradigm-Driven Analogy When looked at from a multilingual perspective, grapheme-to-phoneme conversion is a challenging task, fraught with most of the classical NLP "vexed questions": bottle-neck problem of data acquisition, pervasiveness of exceptions, difficulty to state range and order of rule application, proper treatment of context-sensitive phenomena and long-distance dependencies, and so on. The hand-crafting of transcription rules by a human expert is onerous and time-consuming, and yet, for some European languages, still stops short of a level of correctness and accuracy acceptable for practical applications. We illustrate here a self-learning multilingual system for analogy-based pronunciation which was tested on Italian, English and French, and whose performances are assessed against the output of both statistically and rule-based transcribers. The general point is made that analogy-based self-learning techniques are no longer just psycholinguistically-plausible models, but competitive tools, combining the advantages of using language-independent, self-learning, tractable algorithms, with the welcome bonus of being more reliable for applications than traditional text-to-speech systems. ML 192612 Ontobroker: The Very High Idea The World Wide Web (WWW) is currently one of the most important electronic information sources. However, its query interfaces and the provided reasoning services are rather limited. Ontobroker consists of a number of languages and tools that enhance query access and inference service of the WWW. The technique is based on the use of ontologies. Ontologies are applied to annotate web documents and to provide query access and inference service that deal with the semantics of the presented information. In consequence, intelligent brokering services for web documents can be achieved without requiring to change the semiformal nature of web documents. Introduction The World Wide Web (WWW) contains huge amounts of knowledge about almost all subjects you can think of. HTML documents enriched by multi-media applications provide knowledge in different representations (i.e., text, graphics, animated pictures, video, sound, virtual reality, etc.). Hypertext links between web documents represent r... IR 445880 Market Protocols for Decentralized Supply Chain Formation In order to effectively respond to changing market conditions, business partners must be able to rapidly form supply chains. This thesis approaches the problem of automating supply chain formation---the process of determining the participants in a supply chain, who will exchange what with whom, and the terms of the exchanges---within an economic framework. In this thesis, supply chain formation is formalized as task dependency networks. This model captures subtask decomposition in the presence of resource contention---two important and challenging aspects of supply chain formation. In order to form supply chains in a decentralized fashion, price systems provide an economic framework for guiding the decisions of self-interested agents. In competitive price equilibrium, agents choose optimal allocations with respect to prices, and outcomes are optimal overall. Approximate competitive equilibria yield approximately optimal allocations. Different market protocols are proposed for agents to negotiate the allocation of resources to form supply chains. In the presence of resource contention, these protocols produce better solutions than the greedy protocols common in the artificial intelligence and multiagent systems literature. The first protocol proposed is based on distributed, progressive, price-based auctions, and is analyzed with non-strategic, agent bidding policies. The protocol often converges to high-value supply chains, and when competitive equilibria exist, typically to approximate competitive equilibria. However, complemen- tarities in agent production technologies can cause the protocol to wastefully allocate inputs to agents that do not produce their out... Agents spelt99theorem A Theorem Prover-Based Analysis Tool for Object-Oriented Databases We present a theorem-prover based analysis tool for object-oriented database systems with integrity constraints. Object-oriented database specifications are mapped to higher-order logic (HOL). This allows us to reason about the semantics of database operations using a mechanical theorem prover such as Isabelle or PVS. The tool can be used to verify various semantics requirements of the schema (such as transaction safety, compensation, and commutativity) to support the advanced transaction models used in workflow and cooperative work. We give an example of method safety analysis for the generic structure editing operations of a cooperative authoring system. 1 Introduction Object-oriented specification methodologies and object-oriented programming have become increasingly important in the past ten years. Not surprisingly, this has recently led to an interest in object-oriented program verification in the theorem prover community, mainly using higher-order logic (HOL). Several dif... DB koskela00picsom The PicSOM Retrieval System: Description and Evaluations We have developed an experimental system called PicSOM for retrieving images similar to a given set of reference images in large unannotated image databases. The technique is based on a hierarchical variant of the Self-Organizing Map (SOM) called the Tree Structured Self-Organizing Map (TS-SOM). Given a set of reference images, PicSOM is able to retrieve another set of images which are most similar to the given ones. Each TS-SOM is formed using a different image feature representation like color, texture, or shape. A new technique introduced in PicSOM facilitates automatic combination of the responses from multiple TS-SOMs and their hierarchical levels. This mechanism adapts to the user's preferences in selecting which images resemble each other. In this paper, a brief description of the system and a set of methods applicable to evaluating retrieval performance of image retrieval applications are presented. 1 Introduction Content-based image retrieval (CBIR) has been a subject... HCI 37804 Embodied Evolution: Embodying an Evolutionary Algorithm in a Population of Robots We introduce Embodied Evolution (EE) as a methodology for the automatic design of robotic controllers. EE is an evolutionary robotics (ER) technique that avoids the pitfalls of the simulate-and-transfer method, allows the speed-up of evaluation time by utilizing parallelism, and is particularly suited to future work on multi-agent behaviors. In EE, an evolutionary algorithm is distributed amongst and embodied within a population of physical robots that reproduce with one another while situated in the task environment. We have built a population of eight robots and successfully implemented our first experiments. The controllers evolved by EE compare favorably to hand-designed solutions for a simple task. We detail our methodology, report our initial results, and discuss the application of EE to more advanced and distributed robotics tasks. 1. Introduction Our work is inspired by the following vision. A large number of robots freely interact with each other in a shared environment, atte... AI chen99dynamic Dynamic Agents this paper, we shall explain how dynamic behaviors are obtained and utilized through automatic action installation, and inter-agent communication. We also describe intra-agent communication between the carrier-part and the action part of a dynamic agent, and between di#erent actions carried by the same dynamic agent. We shall also discuss three triggering mechanisms for dynamic behavior modi#cation: planning-based, request-driven, and event-based. Agents devanbu99chime CHIME: Customizable Hyperlink Insertion and Maintenance Engine for Software Engineering Environments Source code browsing is an important part of program comprehension. Browsers expose semantic and syntactic relationships (such as between object references and definitions) in GUI-accessible forms. These relationships are derived using tools which perform static analysis on the original software documents. Implementing such browsers is tricky. Program comprehension strategies vary, and it is necessary to provide the right browsing support. Analysis tools to derive the relevant crossreference relationships are often difficult to build. Tools to browse distributed documents require extensive coding for the GUI, as well as for data communications. Therefore, there are powerful motivations for using existing static analysis tools in conjunction with WWW technology to implement browsers for distributed software projects. The chime framework provides a flexible, customizable platform for inserting HTML links into software documents using information generated by existing software analysis tools. Using the chime specification language, and a simple, retargetable database interface, it is possible to quickly incorporate a range of different link insertion tools for software documents, into an existing, legacy software development environment. This enables tool builders to offer customized browsing support with a well-known GUI. This paper describes the chime architecture, and describes our experience with several re-targeting efforts of this system. 1 DB mihalcea98word Word Sense Disambiguation based on Semantic Density This paper presents a Word Sense Disambiguation method based on the idea of semantic density between words. The disambiguation is done in the context of WordNet. The Internet is used as a raw corpora to provide statistical information for word associations. A metric is introduced and used to measure the semantic density and to rank all possible combinations of the senses of two words. This method provides a precision of 58% in indicating the correct sense for both words at the same time. The precision increases as we consider more choices: 70% for top two ranked and 73% for top three ranked. 1 Introduction Word Sense Disambiguation (WSD) is an open problem in Natural Language Processing. Its solution impacts other tasks such as discourse, reference resolution, coherence, inference and others. WSD methods can be broadly classified into three types: 1. WSD that make use of the information provided by machine readable dictionaries (Cowie et al.1992), (Miller et al.1994), (Agirre and Rig... AI lagoze01open The Open Archives Initiative: Building a low-barrier interoperability framework The Open Archives Initiative (OAI) develops and promotes interoperability solutions that aim to facilitate the efficient dissemination of content. The roots of the OAI lie in the E-Print community. Over the last year its focus has been extended to include all content providers. This paper describes the recent history of the OAI – its origins in promoting E-Prints, the broadening of its focus, the details of its technical standard for metadata harvesting, the applications of this standard, and future plans. Categories and Subject Descriptors IR neti00perceptual Perceptual Interfaces For Information Interaction: Joint Processing Of Audio And Visual Information For Human-Computer Interaction We are exploiting the human perceptual principle of sensory integration (the joint use of audio and visual information) to improve the recognition of human activity (speech recognition, speech event detection and speaker change), intent (intent to speak) and human identity (speaker recognition), particularly in the presence of acoustic degradation due to noise and channel. In this paper, we present experimental results in a variety of contexts that demonstrate the benefit of joint audio-visual processing. HCI pouwelse99feasible A Feasible Low-Power Augmented-Reality Terminal This paper studies the requirements for a truly wearable augmented-reality (AR) terminal. The requirements translate into a generic hardware architecture consisting of programmable modules communicating through a central interconnect. Careful selection of low-power components shows that it is feasible to construct an AR terminal that weighs about 2 kg and roughly dissipates 26 W. With stateof -the-art batteries and a 50% average resource utilization, the terminal can operate for about 10 hours. 1. Introduction The goal of ubiquitous computing is to have computers act as "human assistants" that support us instantly. Computers should move out of our awareness instead of being at the center of our attention [14]. For ubiquitous computing to become reality we need two important technologies to mature: wireless communication (wearability) and augmented reality (user-interface). Wireless communication is obviously required to obtain services provided by an arbitrary computer regardless the ... HCI chen98learningbased Learning-Based Vision and Its Application to Autonomous Indoor Navigation Learning-Based Vision and Its Application to Autonomous Indoor Navigation By Shaoyun Chen Adaptation is critical to autonomous navigation of mobile robots. Many adaptive mechanisms have been implemented, ranging from simple color thresholding to complicated learning with artificial neural networks (ANN). The major focus of this thesis lies in machine learning for vision-based navigation. Two well known vision-based navigation systems are ALVINN and ROBIN developed by Carnegie-Mellon University and University of Maryland, respectively. ALVINN uses a two-layer feedforward neural network while ROBIN relies on a radial basis function network (RBFN). Although current ANN-based methods have achieved great success in vision-based navigation, they have two major disadvantages: (1) Local minimum problem: The training of either multilayer perceptron or radial basis function network can get stuck at poor local minimums. (2) The flexibility problem: After the system has been trained in certain r... ML 400733 View-independent Recognition of Hand Postures Since human hand is highly articulated and deformable, hand posture recognition is a challenging example in the research of view-independent object recognition. Due to the difficulties of the modelbased approach, the appearance-based learning approach is promising to handle large variation in visual inputs. However, the generalization of many proposed supervised learning methods to this problem often suffers from the insufficiency of labeled training data. This paper describes an approach to alleviate this difficulty by adding a large unlabeled training set. Combining supervised and unsupervised learning paradigms, a novel and powerful learning approach, the Discriminant-EM (D-EM) algorithm, is proposed in this paper to handle the case of small labeled training set. Experiments show that D-EM outperforms many other learning methods. Based on this approach, we implement a gesture interface to recognize a set o... IR lau99programming Programming by Demonstration: An Inductive Learning Formulation Although Programming by Demonstration (PBD) has the potential to improve the productivity of unsophisticated users, previous PBD systems have used brittle, heuristic, domain-specific approaches to execution-trace generalization. In this paper we define two applicationindependent methods for performing generalization that are based on well-understood machine learning technology. TGen vs uses version-space generalization, and TGen foil is based on the FOIL inductive logic programming algorithm. We analyze each method both theoretically and empirically, arguing that TGen vs has lower sample complexity, but TGen foil can learn a much more interesting class of programs. Keywords Programming by Demonstration, machine learning, inductive logic programming, version spaces INTRODUCTION Computer users are largely unable to customize massproduced applications to fit their individual tasks. This problem of end-user customization has been addressed in several ways. ffl Adjustable preference... ML kruschwitz01exploiting Exploiting Structure for Intelligent Web Search Together with the rapidly growing amount of online data we register an immense need for intelligent search engines that access a restricted amount of data as found in intranets or other limited domains. This sort of search engines must go beyond simple keyword indexing/matching, but they also have to be easily adaptable to new domains without huge costs. This paper presents a mechanism that addresses both of these points: first of all, the internal document structure is being used to extract concepts which impose a directorylike structure on the documents similar to those found in classified directories. Furthermore, this is done in an efficient way which is largely language independent and does not make assumptions about the document structure. IR knoblock00accurately Accurately and Reliably Extracting Data from the Web: A Machine Learning Approach A critical problem in developing information agents for the Web is accessing data that is formatted for human use. We have developed a set of tools for extracting data from web sites and transforming it into a structured data format, such as XML. The resulting data can then be used to build new applications without having to deal with unstructured data. The advantages of our wrapping technology over previous work are the the ability to learn highly accurate extraction rules, to verify the wrapper to ensure that the correct data continues to be extracted, and to automatically adapt to changes in the sites from which the data is being extracted. 1 Introduction There is a tremendous amount of information available on the Web, but much of this information is not in a form that can be easily used by other applications. There are hopes that XML will solve this problem, but XML is not yet in widespread use and even in the best case it will only address the problem within application domains... IR oard01clef The CLEF 2003 Interactive Track The CLEF 2003 Interactive Track (iCLEF) was the third year of a shared experiment design to compare strategies for cross-language search assistance. Two kinds of experiments were performed: a) experiments in Cross-Language Document Selection, where the user task is to scan a ranked list of documents written in a foreign language, selecting those which seem relevant to a given query. The aim here is to compare di#erent translation strategies for an "indicative" purpose; and b) Full Cross-Language Search experiments, where the user task is to maximize the number of relevant documents that can be found in a foreignlanguage collection with the help of an end-to-end cross-language search system. Participating teams could choose to focus on any aspects of the search task (e.g., query formulation, query translation and/or relevance feedback). This paper describes the shared experiment design and briefly summarizes the experiments run by the five teams that participated. HCI 40032 Rotational Polygon Containment and Minimum Enclosure An algorithm and a robust floating point implementation is given for rotational polygon containment:given polygons P 1 ,P 2 ,P 3 ,...,P k and a container polygon C, find rotations and translations for the k polygons that place them into the container without overlapping. A version of the algorithm and implementation also solves rotational minimum enclosure: givenaclass C of container polygons, find a container C in C of minimum area for which containment has a solution. The minimum enclosure is approximate: it bounds the minimum area between (1-epsilon)A and A. Experiments indicate that finding the minimum enclosure is practical for k = 2, 3 but not larger unless optimality is sacrificed or angles ranges are limited (although these solutions can still be useful). Important applications for these algorithm to industrial problems are discussed. The paper also gives practical algorithms and numerical techniques for robustly calculating polygon set intersection, Minkowski sum, and range in... DB theilmann99disseminating Disseminating Mobile Agents for Distributed Information Filtering An often claimed benefit of mobile agent technology is the reduction of communication cost. Especially the area of information filtering has been proposed for the application of mobile filter agents. However, an effective coordination of agents, which takes into account the current network conditions, is difficult to achieve. This contribution analyses the situation that data distributed among various remote data servers has to be examined with mobile filter agents. We present an approach for coordinating the agents' employment, which minimizes communication costs. Validation studies on the possible cost savings for various constellations show that savings up to 90% can be achieved in the face of actual Internet conditions. 1. Introduction An often claimed benefit of mobile agent technology is the reduction of communication cost, either for decreasing an application's latency or for reducing the load on a network. This prediction has been made especially for scenarios of information... Agents vilalta00quantification A Quantification Of Distance-Bias Between Evaluation Metrics In Classification This paper provides a characterization of bias for evaluation metrics in classification (e.g., Information Gain, Gini, 2 , etc.). Our characterization provides a uniform representation for all traditional evaluation metrics. Such representation leads naturally to a measure for the distance between the bias of two evaluation metrics. We give a practical value to our measure by observing if the distance between the bias of two evaluation metrics correlates with differences in predictive accuracy when we compare two versions of the same learning algorithm that differ in the evaluation metric only. Experiments on real-world domains show how the expectations on accuracy differences generated by the distance-bias measure correlate with actual differences when the learning algorithm is simple (e.g., search for the best single-feature or the best single-rule). The correlation, however, weakens with more complex algorithms (e.g., learning decision trees). Our results sh... ML ng98reconfiguring On Reconfiguring Query Execution Plans in Distributed Object-Relational DBMS Massive database sizes and growing demands for decision support and data mining result in long-running queries in extensible Object-Relational DBMS, particularly in decision support and data warehousing analysis applications. Parallelization of query evaluation is often required for acceptable performance. Yet queries are frequently processed suboptimally due to (1) only coarse or inaccurate estimates of the query characteristics and database statistics available prior to query evaluation; (2) changes in system configuration and resource availability during query evaluation. In a distributed environment, dynamically reconfiguring query execution plans (QEPs), which adapts QEPs to the environment as well as the query characteristics, is a promising means to significantly improve query evaluation performance. Based on an operator classification, we propose an algorithm to coordinate the steps in a reconfiguration and introduce alternatives for execution context checkpointing and restorin... DB straccia01reasoning Reasoning within Fuzzy Description Logics Description Logics (DLs) are suitable, well-known, logics for managing structured knowledge. They allow reasoning about individuals and well defined concepts, i.e. set of individuals with co#hfiP pro# erties. The experience in using DLs inapplicatio#& has sho wn that in many cases we wo#6H like to extend their capabilities. In particular, their use in the co# texto# Multimedia Info#mediafi Retrieval (MIR) leadsto the co# vincement that such DLssho#PF allo w the treatmento f the inherentimprecisio# in multimediao# ject co# tent representatio# and retrieval. In this paper we will present a fuzzyextensio# ALC,co# bining Zadeh's fuzzy lo#zy with a classical DL. In particular,co#rticu beco#FK fuzzy and, thus,reaso#HO6 ab o#fi impreciseco#recis is suppo#ppfi6 We will define its syntax, its semantics, describe its pro# erties and present a co#PHOSfi9 tpro#F&fi9KFS calculus for reasoning in it. ML 346939 Feature Selection in Web Applications By ROC Inflections and Powerset Pruning coetzee,compuman,lawrence,giles¥ A basic problem of information processing is selecting enough features to ensure that events are accurately represented for classification problems, while simultaneously minimizing storage and processing of irrelevant or marginally important features. To address this problem, feature selection procedures perform a search through the feature power set to find the smallest subset meeting performance requirements. Major restrictions of existing procedures are that they typically explicitly or implicitly assume a fixed operating point, and make limited use of the statistical structure of the feature power set. We present a method that combines the Neyman-Pearson design procedure on finite data, with the directed set structure of the Receiver Operating Curves on the feature subsets, to determine the maximal size of the feature subsets that can be ranked in a given problem. The search can then be restricted to the smaller subsets, resulting in significant reductions in computational complexity. Optimizing the overall Receiver Operating Curve also allows for end users to select different operating points and cost functions to optimize. The algorithm also produces a natural method of Boolean representation of the minimal feature combinations that best describe the data near a given operating point. These representations are especially appropriate when describing data using common text-related features useful on the web, such as thresholded TFIDF data. We show how to use these results to perform automatic Boolean query modification generation for distributed databases, such as niche metasearch engines. 1 IR schmidt01xml The XML Benchmark Project With standardization efforts of a query language for XML documents drawing to a close, researchers and users increasingly focus their attention on the database technology that has to deliver on the new challenges that the sheer amount of XML documents produced by applications pose to data management: validation, performance evaluation and optimization of XML query processors are the upcoming issues. Following a long tradition in database research, the XML Store Benchmark Project provides a framework to assess an XML database's abilities to cope with a broad spectrum of different queries, typically posed in real-world application scenarios. The benchmark is intended to help both implementors and users to compare XML databases independent of their own, specific application scenario. To this end, the benchmark o ers a set queries each of which is intended to challenge a particular primitive of the query processor or storage engine. The overall workload wepropose consists of a scalable document database and a concise, yet comprehensive set of queries, which covers the major aspects of query processing. The queries' challenges range from stressing the textual character of the document to data analysis queries, but include also typical ad-hoc queries. We complement our research with results obtained from running the benchmark on our XML database platform. They are intended to give a rst baseline, illustrating the state of the art. DB 298827 Audio Signal Classification Audio signal classification (ASC) consists of extracting relevant features from a sound, and of using these features to identify into which of a set of classes the sound is most likely to fit. The feature extraction and grouping algorithms used can be quite diverse depending on the classification domain of the application. This paper presents background necessary to understand the general research domain of ASC, including signal processing, spectral analysis, psychoacoustics and auditory scene analysis. Also presented are the basic elements of classification systems. Perceptual and physical features are discussed, as well as clustering algorithms and analysis duration. Neural nets and hidden Markov models are discussed as they relate to ASC. These techniques are presented with an overview of the current state of the ASC research literature. 1 Introduction This paper will present a review of the state of the current research literature pertaining to audio signal classifiation (ASC). ... AI pitoura99scalable Scalable Processing of Read-Only Transactions in Broadcast Push Recently, push-based delivery has attracted considerable attention as a means of disseminating information to large client populations in both wired and wireless settings. In this paper, we address the problem of ensuring the consistency and currency of client read-only transactions in the presence of updates. To this end, additional control information is broadcast. A suite of methods is proposed that vary in the complexity and volume of the control information transmitted and subsequently differ in response times, degrees of concurrency, and space and processing overheads. The proposed methods are combined with caching to improve query latency. The relative advantages of each method are demonstrated through both simulation results and qualitative arguments. Read-only transactions are processed locally at the client without contacting the server and thus the proposed approaches are scalable, i.e., their performance is independent of the number of clients. 1. Introduction In traditio... DB 2056 Interaction between Path and Type Constraints XML [7], which is emerging as an important standard for data exchange on the World-Wide Web, highlights the importance of semistructured data. Although the XML standard itself does not require any schema or type system, a number of proposals [6, 17, 19] have been developed that roughly correspond to data definition languages. These allow one to constrain the structure of XML data by imposing a schema on it. These and other proposals also advocate the need for integrity constraints, another form of constraints that should, for example, be capable of expressing inclusion constraints and inverse relationships. The latter have recently been studied as path constraints in the context of semistructured data [4, 9]. It is likely that future XML proposals will involve both forms of constraints, and it is therefore appropriate to understand the interaction between them. This paper investigates that interaction. In particular it studies constraint implication problems, which are important both i... DB shu98using Using Constraint Satisfaction for View Update Translation View update is the problem of translating update requests against a view into update requests against the base data. In this paper, we present a novel approach to this problem in relational databases. Using conditional tables to represent relational views, we translate a view update problem into a disjunction of a number of constraint satisfaction problems (CSPs). Solutions to the CSPs correspond to possible translations of the view update. Semantic information to resolve ambiguity can be handled as additional constraints in the CSPs. This approach enables us to apply the rich results of the CSP research to analyze and solve an important problem in database management. DB bates98state The State of the Art in Distributed and Dependable Computing This report is dedicated to the memory of Henrique Fonseca HCI arpinar00open An Open Electronic Marketplace through Agent-based Workflows: MOPPET We propose an electronic marketplace architecture, called MOPPET, where the commerce processes in the marketplace are modeled as adaptable agent-based workflows. The higher level of abstraction provided by the workflow technology makes the customization of electronic commerce processes for different users possible. Agent-based implementation, on the other hand, provides for a highly reusable component-based workflow architecture as well as negotiation ability and the capability to adapt to dynamic changes in the environment. Agent communication is handled through Knowledge Query and Manipulation Language (KQML). A workflow-based architecture also makes it possible for complete modeling of electronic commerce processes by allowing involved parties to be able to invoke already existing applications or to define new tasks and to restructure the control and data flow among the tasks to create custom built process definitions. In the proposed architecture all data exchanges are realized through Extensible Markup Language (XML) providing uniformity, simplicity and a highly open and interoperable architecture. Metadata of activities are expressed through Resource Description Framework (RDF). Common Business Library (CBL) is used for achieving interoperability across business domains and domain specific Document Type Definitions (DTDs) are used for vertical industries. We provide our own specifications for missing DTDs to be replaced by the original specifications when they become available. IR 502274 Interpretation of Shape-related Iconic Gestures in Virtual Environments The interpretation of iconic gestures in spatial domains is a promising idea to improve the communicative capabilities of human-computer interfaces. So far, approaches towards gesture recognition focused mainly on deictic and emblematic gestures. Iconics, viewed as iconic signs in the sense of Peirce, are different from deictics and emblems, for their relation to the referent is based on similarity. In the work reported here, the breakdown of the complex notion of similarity provides the key idea towards a computational model of gesture semantics for iconic gestures. Based on an empirical study,we describe first steps towards a recognition model for shape-related iconic gestures and its implementation in a prototype gesture recognition system. Observations are focused on spatial concepts and their relation to features of iconic gestural expressions. The recognition model is based on a graphmatching method which compares the decomposed geometrical structures of gesture and object. HCI sanchez98agora Agora: Enhancing Group Awareness and Collaboration in Floristic Digital Libraries . Digital libraries can be regarded as virtual spaces in which collaborative scholarly research can be conducted. Floristic digital libraries provide such collaboration spaces for scientists working on solutions for Earth's biodiversity problems. However, group awareness and collaboration are not easily achieved in an increasingly distributed environment such as the virtual space in which digital library users (particularly botanists and biologists) do their work. We describe an environment that enables group awareness, communication, and collaboration among users in a globally accessible floristic digital library. This is achieved by extending existing library facilities with recommendation and alerting services, as well as various communication interfaces. Keywords: Recommendation services, group awareness, agents, floristic digital libraries. 1. Introduction Digital libraries comprise highly complex and dynamic information spaces on top of which a variety of services are provide... IR wang01supporting Supporting Workspace Awareness in Distance Learning Environments: Issues and Experiences in the Development of a Collaborative Learning System In recent years, we have witnessed an enormous growth in networks and related technologies. Course materials are increasingly published on web servers, and students are encouraged to access these at leisure. Distance learning via the WWW shifted the education paradigm from teacher-centered instruction to user-centered collaborative learning. Systems that allow users to learn collaboratively are increasingly interesting to scientific communities and learning organizations. We initially designed and prototyped a collaborative system to support collaborative learning over the Internet. A usability study of the first prototype revealed the importance of awareness information. Our review of three wellknown collaborative systems finds that such systems today also lack support for awareness information, especially workspace awareness. We then consider various types of awareness in collaborative learning situations and set out the design requirements of our system. From these requirements, we have designed and prototyped several awareness widgets for a typical collaborative tool: the shared electronic whiteboard. These widgets help learners maintain awareness of other learners' interactions with the shared workspace. HCI 353348 A Formal Approach to Detecting Security Flaws in Object-Oriented Databases This paper adopts the method-based authorization model and assumes the following database management policies. Let (m, (c 1 ,c 2 , ...,c n )) be in an authorization for a user u. DB khudanpur99maximum A Maximum Entropy Language Model Integrating N-Grams And Topic Dependencies For Conversational Speech Recognition A compact language model which incorporates local dependencies in the form of N-grams and long distance dependencies through dynamic topic conditional constraints is presented. These constraints are integrated using the maximum entropy principle. Issues in assigning a topic to a test utterance are investigated. Recognition results on the Switchboard corpus are presented showing that with a very small increase in the number of model parameters, reduction in word error rate and language model perplexity are achieved over trigram models. Some analysis follows, demonstrating that the gains are even larger on content-bearing words. The results are compared with those obtained by interpolating topicindependent and topic-specific N-gram models. The framework presented here extends easily to incorporate other forms of statistical dependencies such as syntactic word-pair relationships or hierarchical topic constraints. 1. INTRODUCTION Language modeling is a crucial component of systems that c... IR bellavista01how How to Monitor and Control Resource Usage in Mobile Agent Systems The Mobile Agent technology has already shown its advantages, but at the same time has already remarked new problems currently limiting its diffusion in commercial environments. A key issue is to control the operations that foreign mobile agents are authorized to perform on hosting execution environments. It is necessary not only to rule the MA access to resources but also to control resource usage of admitted agents at execution time, for instance to protect against possible denial-of-service attacks. The paper presents a solution framework for the on-line monitoring and control of Java-based MA platforms. In particular, it describes the design and implementation of MAPI, an on-line monitoring component that we have integrated within the SOMA system. The paper shows how to use MAPI as the building block of a distributed monitoring tool that gives application- and kernel-level information about the state of mobile agents and their resource usage, thus enabling the enforcement of management policies on MA resource consumption. 1. Agents mchugh99optimizing Optimizing Branching Path Expressions Path expressions form the basis of most query languages for semistructured data and XML, specifying traversals through graph-based data. We consider the query optimization problem for path expressions that "branch," or specify traversals through two or more related subgraphs; such expressions are common in nontrivial queries over XML data. Searching the entire space of query plans for branching path expressions is generally infeasible, so we introduce several heuristic algorithms and postoptimizations that generate query plans for branching path expressions. All of our algorithms have been implemented in the Lore database system for XML, and we report experimental results over a variety of database and query shapes. We compare optimization and execution times across our suite of algorithms and post-optimizations, and for small queries we compare against the optimal plan produced by an exhaustive search of the plan space. 1 Introduction Wo r k i n semistructured data [Abi97, ... DB 452812 Pedagogical Content Knowledge in a Tutorial Dialogue System to Support Self-Explanation We are engaged in a research project to create a tutorial dialogue system that helps students learn through self-explanation. Our current prototype is able to analyze students' general explanations of their problem-solving steps, stated in their own words, recognize the types of omissions that we often see in these explanations, and provide feedback. Our approach to architectural tradeoffs is to equip the system with a sophisticated NLU component but to keep dialogue management simple. The system has a knowledge-based NLU component, which performed with 81% accuracy in a preliminary evaluation study. The system's approach to dialogue management can be characterised as "classify and react". In each dialogue cycle, the system classifies the student input with respect to a hierarchy of explanation categories that represent common ways of stating complete or incomplete explanations of geometry rules. The system then provides feedback based on that classification. We consider what extensions are necessary or desirable in order to make the dialogues more robust. HCI renambot00cavestudy CAVEStudy: an Infrastructure for Computational Steering in Virtual Reality Environments We present the CAVEStudy system that enables scientists to interactively steer a simulation from a virtual reality (VR) environment. No modification to the source code is necessary. CAVEStudy allows interactive and immersive analysis of a simulation running on a remote computer. Using a high-level description of the simulation, the system generates the communication layer (based on CAVERNSoft) needed to control the execution and to gather data at runtime. We describe three case-studies implemented with CAVEStudy: soccer simulation, diode laser simulation, and molecular dynamics. 1. Introduction High-speed networks and high performance graphics open opportunities for completely new types of applications. As a result, the world of scientific computing is moving away from the batch-oriented management to interactive programs. Also, virtual reality (VR) systems are now commercially available, but so far scientists mainly use them for off-line visualization of data sets produced by a simu... HCI traum99speech Speech Acts for Dialogue Agents this paper by the U.S. Army Research Office under contract/grant number DAAH 04 95 10628 and the U.S. National Science Foundation under grant IRI9311988. Some of the work described above was developed in collaboration with James Allen and supported by ONR/DARPA under grant number N00014-92J -1512, by ONR under research grant number N00014-90-J-1811, and by NSF under grant number IRI-9003841. Agents amyot00extension On the Extension of UML with Use Case Maps Concepts . Descriptions of reactive systems focus heavily on behavioral aspects, often in terms of scenarios. To cope with the increasing complexity of services provided by these systems, behavioral aspects need to be handled early in the design process with flexible and concise notations as well as expressive concepts. UML offers different notations and concepts that can help describe such services. However, several necessary concepts appear to be absent from UML, but present in the Use Case Map (UCM) scenario notation. In particular, Use Case Maps allow scenarios to be mapped to different architectures composed of various component types. The notation supports structured and incremental development of complex scenarios at a high level of abstraction, as well as their integration. UCMs specify variations of run-time behavior and scenario structures through sub-maps "pluggable" into placeholders called stubs. This paper presents how UCM concepts could be used to extend the semantics... DB mohan99repeating Repeating History beyond ARIES In this paper, I describe first the background behind the development of the original ARIES recovery method, and its significant impact on the commercial world and the research community. Next, I provide a brief introduction to the various concurrency control and recovery methods in the ARIES family of algorithms. Subsequently, I discuss some of the recent developments affecting the transaction management area and what these mean for the future. In ARIES, the concept of repeating history turned out to be an important paradigm. As I examine where transaction management is headed in the world of the internet, I observe history repeating itself in the sense of requirements that used to be considered significant in the mainframe world (e.g., performance, availability and reliability) now becoming important requirements of the broader information technology community as well. 1. Introduction Transaction management is one of the most important functionalities provided by a... DB frank98generating Generating Accurate Rule Sets Without Global Optimization The two dominant schemes for rule-learning, C4.5 and RIPPER, both operate in two stages. First they induce an initial rule set and then they refine it using a rather complex optimization stage that discards (C4.5) or adjusts (RIPPER) individual rules to make them work better together. In contrast, this paper shows how good rule sets can be learned one rule at a time, without any need for global optimization. We present an algorithm for inferring rules by repeatedly generating partial decision trees, thus combining the two major paradigms for rule generation—creating rules from decision trees and the separate-and-conquer rule-learning technique. The algorithm is straightforward and elegant: despite this, experiments on standard datasets show that it produces rule sets that are as accurate as and of similar size to those generated by C4.5, and more accurate than RIPPER’s. Moreover, it operates efficiently, and because it avoids postprocessing, does not suffer the extremely slow performance on pathological example sets for which the C4.5 method has been criticized. ML kurimo99latent Latent Semantic Indexing by Self-Organizing Map An important problem for the information retrieval from spoken documents is how to extract those relevant documents which are poorly decoded by the speech recognizer. In this paper we propose a stochastic index for the documents based on the Latent Semantic Analysis (LSA) of the decoded document contents. The original LSA approach uses Singular Value Decomposition to reduce the dimensionality of the documents. As an alternative, we propose a computationally more feasible solution using Random Mapping (RM) and Self-Organizing Maps (SOM). The motivation for clustering the documents by SOM is to reduce the effect of recognition errors and to extract new characteristic index terms. Experimental indexing results are presented using relevance judgments for the retrieval results of test queries and using a document perplexity defined in this paper to measure the power of the index models. 1. INTRODUCTION In this paper we present methods for indexing speech data which has been automatically t... IR huget02desiderata Desiderata for Agent Oriented Programming Languages Multiagent system designers need programming languages in order to develop agents and multiagent systems. Current approaches consist to use classical programming languages like C or C++ and above all Java which is the most preferred language by agent community thanks to its rich library of functions. The aim of Java is not to design multiagent systems so it does not encompass multiagent features. The aim of this paper is to present a set of characteristics which could be present in an agent-oriented programming language. This paper also describes what kind of multiagent systems could be developed with this set of characteristics. Agents jenkins00primitivebased Primitive-Based Movement Classification for Humanoid Imitation . Motor control is a complex problem and imitation is a powerful mechanism for acquiring new motor skills. In this paper, we describe perceptuo-motor primitives, a biologically-inspired notion for a basis set of perceptual and motor routines. Primitives serve as a vocabulary for classifying and imitating observed human movements, and are derived from the imitator's motor repertoire. We describe a model of imitation based on such primitives and demonstrate the feasibility of the model in a constrained implementation. We present approximate motion reconstruction generated from visually captured data of typically imitated tasks taken from aerobics, dancing, and athletics. 1 Introduction Imitation is a powerful mechanism for acquiring new skills. It involves an intricate interaction between perceptual and motor mechanisms, both of which are complex in themselves. Research into vision and motor control has explored the role of "subroutines", schemas [1], and other variants based on ... ML aha97learning Learning to Refine Case Libraries: Initial Results . Conversational case-based reasoning (CBR) systems, which incrementally extract a query description through a user-directed conversation, are advertised for their ease of use. However, designing large case libraries that have good performance (i.e., precision and querying efficiency) is difficult. CBR vendors provide guidelines for designing these libraries manually, but the guidelines are difficult to apply. We describe an automated inductive approach that revises conversational case libraries to increase their conformance with design guidelines. Revision increased performance on three conversational case libraries. 1 Introduction In the context of the ECML-97 Workshop entitled Case-Based Learning: Beyond Classification of Feature Vectors, this paper's contribution focuses on using machine learning methods to assist in the design of case libraries. These libraries are designed for solution retrieval rather than classification tasks, and each case might contain a unique solution. Cas... ML murray01specifying Specifying Agents with UML in Robotic Soccer The use of agents and multiagent systems is widespread in computer science nowadays. Thus the need for methods to specify agents in a clear and simple manner arises. In this paper we propose an approach to specifying agents with the help of UML statecharts. Agents are specified on different levels of abstraction. In addition a method for specifying multiagent plans with explicit cooperation is shown. As an example domain we chose robotic soccer, which lays the basis of the annual RoboCup competitions. Robotic soccer is an ideal testbed for research in the fields of robotics and multiagent systems. In the RoboCup Simulation League the research focus is laid on agents and multiagent systems, and we will demonstrate our approach by using examples from this domain. Keywords: Multiagent Systems, Unified Modeling Language (UML), Specification, RoboCup, Robotic Soccer 1 Agents 48156 Agents in Annotated Worlds Virtual worlds offer great potential as environments for education, entertainment, and collaborative work. Agents that function effectively in heterogeneous virtual spaces must have the ability to acquire new behaviors and useful semantic information from those contexts. The human-computer interaction literature discusses how to construct spaces and objects that provide "knowledge in the world" that aids human beings to perform these tasks. In this paper, we describe how to build comparable annotated environments containing explanations of the purpose and uses of spaces and activities that allow agents quickly to become intelligent actors in those spaces. Examples are provided from our application domain, believable agents acting as inhabitants and guides in a children's exploratory world. Keywords: believability, human-like qualities of synthetic agents, synthetic agents 1. Introduction Today's virtual environments present opportunities for simulation and interaction involving ma... Agents weber00unsupervised Unsupervised Learning of Models for Recognition Abstract. We present a method to learn object class models from unlabeled and unsegmented cluttered scenes for the purpose of visual object recognition. We focus on a particular type of model where objects are represented as flexible constellations of rigid parts (features). The variability within a class is represented by a joint probability density function (pdf) on the shape of the constellation and the output of part detectors. In a first stage, the method automatically identifies distinctive parts in the training set by applying a clustering algorithm to patterns selected by an interest operator. It then learns the statistical shape model using expectation maximization. The method achieves very good classification results on human faces and rear views of cars. 1 Introduction and Related Work We are interested in the problem of recognizing members of object classes, where we define an object class as a collection of objects which share characteristic features or parts that are visually similar and occur in similar spatial configurations. When building models for object classes of this type, one is faced with three problems (see Fig. 1). ML wiederhold95valueadded Value-added Mediation in Large-Scale Information Systems Many information-processing tasks can be part of multiple customer applications, as summarizing stock prices, integrating catolog information from several companies in the same line of business, predicting the weather, and checking on transportation resources. We assign such sharable services to an active middleware layer, interposed between clients and servers. We define domain-specific mediator modules to populate this layer. Such mediating services must be of value to the customers, so that it will benefit their client applications to access mediators rather than the server sources directly. Several categories of value can be considered: improvement in access and coverage, improvement of content, and delegation of maintenance. We will define criteria for mediating modules: ownership by party who assumes responsibility for the rseults of the services, domain-specificity to delimit the scope of such a responsibility, and, of course, conformance with interface standards that ... HCI jansche01information Information Extraction via Heuristics for a Movie Showtime Query System Semantic interpretation for limited-domain spoken dialogue systems often amounts to extracting information from utterances. For a system that provides movie showtime information, queries are classified along four dimensions: question type, and movie titles, towns and theaters that were mentioned. Simple heuristics suffice for constructing highly accurate classifiers for the latter three attributes; classifiers for the question type attribute are induced from data using features tailored to spoken language phenomena. Since separate classifiers are used for the four attributes, which are not independent, certain errors can be detected and corrected, thus increasing robustness. 1. IR ha98geometric Geometric Foundations for Interval-Based Probabilities The need to reason with imprecise probabilities arises in a wealth of situations ranging from pooling of knowledge from multiple experts to abstraction-based probabilistic planning. Researchers have typically represented imprecise probabilities using intervals and have developed a wide array of different techniques to suit their particular requirements. In this paper we provide an analysis of some of the central issues in representing and reasoning with interval probabilities. At the focus of our analysis is the probability cross-product operator and its interval generalization, the cc-operator. We perform an extensive study of these operators relative to manipulation of sets of probability distributtions. This study provides insight into the sources of the strengths and weaknesses of various approaches to handling probability intervals. We demonstrate the application of our results to the problems of inference in interval Bayesian networks and projection and evaluation of abstract pro... AI handschuh02authoring Authoring and Annotation of Web Pages in CREAM Richly interlinked, machine-understandable data constitute the basis for the Semantic Web. We provide a framework, CREAM, that allows for creation of metadata. While the annotation mode of CREAM allows to create metadata for existing web pages, the authoring mode lets authors create metadata --- almost for free --- while putting together the content of a page. As a particularity of our framework, CREAM allows to create relational metadata, i.e. metadata that instantiate interrelated definitions of classes in a domain ontology rather than a comparatively rigid template-like schema as Dublin Core. We discuss some of the requirements one has to meet when developing such an ontology-based framework, e.g. the integration of a metadata crawler, inference services, document management and a meta-ontology, and describe its implementation, viz. Ont-O-Mat a component-based, ontology-driven Web page authoring and annotation tool. IR 235157 Face Detection Using Mixtures of Linear Subspaces We present two methods using mixtures of linear subspaces for face detection in gray level images. One method uses a mixture of factor analyzers to concurrently perform clustering and, within each cluster, perform local dimensionality reduction. The parameters of the mixture model are estimated using an EM algorithm. A face is detected if the probability of an input sample is above a predefined threshold. The other mixture of subspaces method uses Kohonen’s self-organizing map for clustering and Fisher Linear Discriminant to find the optimal projection for pattern classification, and a Gaussian distribution to model the class-conditional density function of the projected samples for each class. The parameters of the class-conditional density functions are maximum likelihood estimates and the decision rule is also based on maximum likelihood. A wide range of face images including ones in different poses, with different expressions and under different lighting conditions are used as the training set to capture the variations of human faces. Our methods have been tested on three sets of 225 images which contain 871 faces. Experimental results on the first two datasets show that our methods perform as well as the best methods in the literature, yet have fewer false detects. 1 ML koller98probabilistic Probabilistic Frame-Based Systems Two of the most important threads of work in knowledge representation today are frame-based representation systems (FRS's) and Bayesian networks (BNs). FRS's provide an excellent representation for the organizational structure of large complex domains, but their applicability is limited because of their inability to deal with uncertainty and noise. BNs provide an intuitive and coherent probabilistic representation of our uncertainty, but are very limited in their ability to handle complex structured domains. In this paper, we provide a language that cleanly integrates these approaches, preserving the advantages of both. Our approach allows us to provide natural and compact definitions of probability models for a class, in a way that is local to the class frame. These models can be instantiated for any set of interconnected instances, resulting in a coherent probability distribution over the instance properties. Our language also allows us to represent important types of uncertainty tha... AI barros97towards Towards Real-Scale Business Transaction Workflow Modelling While the specification languages of workflow management systems focus on process execution semantics, the successful development of workflows relies on a fuller conceptualisation of business processing, including process semantics. For this, a wellspring of modelling techniques, paradigms and informal-formal method extensions which address the analysis of organisational processing structures (enterprise modelling) and communication (based on speech-act theory), is available. However, the characterisations - indeed the cognition - of workflows still appears coarse. In this paper, we provide the complementary, empirical insight of a real-scale business transaction workflow. The development of the workflow model follows a set of principles which we believe address workflow modelling suitability. Through the principles, advanced considerations including asynchronous as well as synchronous messaging, temporal constraints and a service-oriented perspective are motivated. By illust... HCI dimitrova02web Web Genre Visualization Web users vary widely in terms of their expertise on the topics for which they search, the amount of detail they seek, etc. Unfortunately, today's one-size-fits-all Web search services do not cater to such individual preferences. For example, it is difficult to query for documents that give extensive detail but assume modest prior expertise. We describe how shallow text classification techniques can be used to sort the documents returned by Web search services according to genre dimensions such as level of expertise and amount of detail, and propose a simple visualization interface that helps users rapidly find appropriate documents. (Keywords: document genre; information retrieval, visualisation, text classification, shallow linguistic processing) Motivation Consider two users seeking information about Pearson' correlation coefficient. Alice is writing a data-analysis program and needs a web page to remind her of the equations. Bob, a teacher, wants to point his pupils to an overview that isn't bogged down in equations. IR horvitz99principles Principles of Mixed-Initiative User Interfaces Recent debate has centered on the relative promise of focusing user-interface research on developing new metaphors and tools that enhance users' abilities to directly manipulate objects versus directing effort toward developing interface agents that provide automation. In this paper, we review principles that show promise for allowing engineers to enhance human---computer interaction through an elegant coupling of automated services with direct manipulation. Key ideas will be highlighted in terms of the LookOut system for scheduling and meeting management. Keywords Intelligent agents, direct manipulation, user modeling, probability, decision theory, UI design INTRODUCTION There has been debate among researchers about where great opportunities lay for innovating in the realm of human--- computer interaction [10]. One group of researchers has expressed enthusiasm for the development and application of new kinds of automated services, often referred to as interface "agents." The effo... HCI edmond98achieving Achieving Workflow Adaptability by means of Reflection Belief in the importance of business processes has triggered considerable interest in the workflow systems that automate these processes. However, of the two competing management philosophies that promulgate business processes, Business Process Reengineering proposes radical change, whereas Continuous Process Improvement places much greater emphasis on adaptability. The former school is somewhat discredited, whereas the latter school seems more likely to endure, thus making the flexibility and evolution of workflows an issue of increasing importance. In this paper, we present a programmable object-oriented metalevel framework which aims to reveal the processes of assembling and coordinating the tasks that make up business processes. This is achieved by isolating four key facets -- state, behaviour, location and coordination. In particular, we open up the general process of task coordination and specification, allowing for extensions in a planned way. By suitable manipulation of coordin... DB 293286 Managing Historical Semistructured Data this article appeared in [6]. DB chen01optimal Optimal Anytime Search For Constrained Nonlinear Programming In this thesis, we study optimal anytime stochastic search algorithms (SSAs) for solving general constrained nonlinear programming problems (NLPs) in discrete, continuous and mixed-integer space. The algorithms are general in the sense that they do not assume differentiability or convexity of functions. Based on the search algorithms, we develop the theory of SSAs and propose optimal SSAs with iterative deepening in order to minimize their expected search time. Based on the optimal SSAs, we then develop optimal anytime SSAs that generate improved solutions as more search time is allowed. Our SSAs for solving general constrained NLPs are based on the theory of discrete con-strained optimization using Lagrange multipliers that shows the equivalence between the set of constrained local minima (CLMdn) and the set of discrete-neighborhood saddle points (SPdn). To implement this theory, we propose a general procedural framework for locating an SPdn. By incorporating genetic algorithms in the framework, we evaluate new constrained search algorithms: constrained genetic algorithm (CGA) and combined constrained simulated annealing and genetic algorithm (CSAGA). AI 351802 Design and Implementation of the OLOG Deductive Object-Oriented Database Management System . OLOG is a novel deductive database system for advanced intelligent information system applications. It directly supports eective storage, ecient access and inference of large amount of persistent data with complex structures. It provides a SQL-like data denition language and data manipulation language, and a declarative rule-based query language. It combines the best of the deductive, object-oriented, and objectrelational approaches in a uniform framework. This paper describes the design and implementation of the OLOG system. 1 Introduction Deductive, object-oriented, and object-relational databases are three important extensions of the traditional relational database technology. Deductive databases stem from the integration of logic programming and relational databases. It oers representational and operational uniformity, reasoning capabilities, recursion, declarative querying, ecient secondary storage access, etc. However, deductive databases based on relational databas... DB 38463 Formalising the Knowledge Content of Case Memory Systems Discussions of case-based reasoning often reflect an implicit assumption that a case memory system will become better informed, i.e. will increase in knowledge, as more cases are added to the case-base. This paper considers formalisations of this `knowledge content' which are a necessary preliminary to more rigourous analysis of the performance of case-based reasoning systems. In particular we are interested in modelling the learning aspects of case-based reasoning in order to study how the performance of a case-based reasoning system changes as it accumlates problem-solving experience. The current paper presents a `case-base semantics' which generalises recent formalisations of case-based classification. Within this framework, the paper explores various issues in assuring that these sematics are well-defined, and illustrates how the knowledge content of the case memory system can be seen to reside in both the chosen similarity measure and in the cases of the case-base. 1 Introduction ... ML 469106 Time Series Classification by Boosting Interval Based Literals A supervised classification method for temporal series, even multivariate, is presented. It is based on boosting very simple classifiers: clauses with one literal in the body. The background predicates are based on temporal intervals. Two types of predicates are used: i) relative predicates, such as "increases" and "stays", and ii) region predicates, such as "always" and "sometime", which operate over regions in the dominion of the variable. Experiments on di#erent data sets, several of them obtained from the UCI repositories, show that the proposed method is highly competitive with previous approaches. Keywords: time series classification, interval based literals, boosting, machine learning. 1 DB wachsmuth01lifelike Lifelike Gesture Synthesis and Timing for Conversational Agents Besides the inclusion of gesture recognition devices as an intuitive input modality, the synthesis of lifelike gesture is finding growing attention in human-computer interface research. In particular, the generation of synthetic gesture in connection with text-to-speech systems is one of the goals for embodied conversational agents which have become a new paradigm for the study of gesture and for human-computer interface [1]. Embodied conversational agents are computer-generated characters that demonstrate similar properties as humans in face-to-face conversation, including the ability to produce and respond to verbal and nonverbal communication. They may represent the computer in an interaction with a human or represent their human users as "avatars " in a computational environment. In this context, this contribution focusses on an approach for synthesizing lifelike gestures for an articulated virtual agent, with particular emphasis on how to achieve temporal coordination with external information such as the signal generated by a text-to-speech system. The context of this research is the conception of an "articulated communicator " that conducts multimodal dialogue with a human partner in cooperating on a construction task. Gesture production and performance in humans is a complex and multi-stage process. HCI roman00legorb LegORB and Ubiquitous CORBA The increasing popularity of ubiquitous computing and the new approaches for low-consumption, shortrange wireless connectivity will enable a future with hundreds of heterogeneous devices interconnected to achieve a common task. However, communication among those network enabled heterogeneous devices requires standard protocols and well defined interfaces. While existing middleware architectures already offer standard mechanisms (DCOM, CORBA, JINI), they are, in most of the cases, not suitable for most of the heterogeneous devices. The resources required by those middleware solutions normally exceed the computational limits of the heterogeneous devices We present in this paper a minimalist component-based Object Request Broker (ORB) that can be dynamically reconfigured and requires, for the smallest configuration, 6Kb of memory. Introduction The incoming ubiquitous computing trend allows the existence of collections of network-enabled devices attached to rooms, people and buildings.... HCI sycara95using Using Case-Based Reasoning to Acquire User Scheduling Preferences that Change over Time Production/Manufacturing scheduling typically involves the acquisition of user optimization preferences. The ill-structuredness of both the problem space and the desired objectives make practical scheduling problems difficult to formalize and costly to solve, especially when problem configurations and user optimization preferences change over time. This paper advocates an incremental revision framework for improving schedule quality and incorporating user dynamically changing preferences through Case-Based Reasoning. Our implemented system, called CABINS, records situation-dependent tradeoffs and consequences that result from schedule revision to guide schedule improvement. The preliminary experimental results show that CABINS is able to effectively capture both user static and dynamic preferences which are not known to the system and only exist implicitly in a extensional manner in the case base. 1 Introduction Scheduling deals with allocation of a limited set of resources to a nu... ML 295535 The Adaptive Agent Architecture: Achieving FaultTolerance Using Persistent Broker Teams Brokers are used in many multi-agent systems for locating agents, for routing and sharing information, for managing the system, and for legal purposes, as independent third parties. However, these multi-agent systems can be incapacitated and rendered non-functional when the brokers become inaccessible due to failures such as machine crashes, network breakdowns, and process failures that can occur in any distributed software system. We propose that the theory of teamwork can be used to create robust brokered architectures that can recover from broker failures, and we present the Adaptive Agent Architecture (AAA) to show the feasibility of this approach. The AAA brokers form a team with a joint commitment to serve any agent that registers with the broker team as long as the agent remains registered with the team. This commitment enables the brokers to substitute for each other when needed. A multiagent system based on the AAA can continue to work despite broker failures as long... Agents 201746 Incremental and Interactive Sequence Mining The discovery of frequent sequences in temporal databases is an important data mining problem. Most current work assumes that the database is static, and a database update requires rediscovering all the patterns by scanning the entire old and new database. In this paper, we propose novel techniques for maintaining sequences in the presence of a) database updates, and b) user interaction (e.g. modifying mining parameters). This is a very challenging task, since such updates can invalidate existing sequences or introduce new ones. In both the above scenarios, we avoid re-executing the algorithm on the entire dataset, thereby reducing execution time. Experimental results confirm that our approach results in substantial performance gains. 1 Introduction Sequence mining is an important data mining task, where one attempts to discover frequent sequences over time, of attribute sets in large databases. This problem was originally motivated by applications in the retailing industry, including... DB 275630 On the Learnability and Design of Output Codes for Multiclass Problems . Output coding is a general framework for solving multiclass categorization problems. Previous research on output codes has focused on building multiclass machines given predefined output codes. In this paper we discuss for the first time the problem of designing output codes for multiclass problems. For the design problem of discrete codes, which have been used extensively in previous works, we present mostly negative results. We then introduce the notion of continuous codes and cast the design problem of continuous codes as a constrained optimization problem. We describe three optimization problems corresponding to three different norms of the code matrix. Interestingly, for the l 2 norm our formalism results in a quadratic program whose dual does not depend on the length of the code. A special case of our formalism provides a multiclass scheme for building support vector machines which can be solved efficiently. We give a time and space efficient algorithm for solving the quadratic program. We describe preliminary experiments with synthetic data show that our algorithm is often two orders of magnitude faster than standard quadratic programming packages. We conclude with the generalization properties of the algorithm. Keywords: Multiclass categorization,output coding, SVM 1. IR lueg00information Information Seeking as Socially Situated Activity this paper, we discuss implications of situatedness in its social and cultural meaning in the context of information seeking support. We proceed as follows. First, we discuss some of the varying meanings of the term \situated". Then, we outline how we interpret \accounting for situatedness" in the context of information seeking support. Finally, we discuss tools that implement aspects of what we consider important in this context. HCI yang98feature Feature Subset Selection Using A Genetic Algorithm Practical pattern classification and knowledge discovery problems require selection of a subset of attributes or features (from a much larger set) to represent the patterns to be classified. This paper presents an approach to the multi-criteria optimization problem of feature subset selection using a genetic algorithm. Our experiments demonstrate the feasibility of this approach for feature subset selection in the automated design of neural networks for pattern classification and knowledge discovery. 1 Introduction Many practical pattern classification tasks (e.g., medical diagnosis) require learning of an appropriate classification function that assigns a given input pattern (typically represented using a vector of attribute or feature values) to one of a finite set of classes. The choice of features, attributes, or measurements used to represent patterns that are presented to a classifier affect (among other things): ffl The accuracy of the classification function that can be learn... ML rizzo97personalitydriven Personality-Driven Social Behaviors in Believable Agents Agents are considered "believable" when they are viewed by an audience as endowed with thoughts, desires, and emotions, typical of different personalities. The paper describes our work in progress aimed at realizing believable agents that perform helping behaviors influenced by their own personalities; the latter are represented as different clusters of prioritized goals and preferences over plans for achieving the goals. The implementation is based on the integration of a statebased planner that serves as the reasoning tool for the agents and a situation-driven execution system. Introduction "There is a notion in the Arts of `believable character '. It does not mean an honest or reliable character, but one that provides the illusion of life, and thus permits the audience's suspension of disbelief 1 . The idea of believability has long been studied and explored in literature, theater, film, radio, drama, and other media ". (Bates 1994) Believability therefore refers to a character's... HCI friedman01multivariate Multivariate Information Bottleneck The Information bottleneck method is an unsupervised non-parametric data organization technique. Given a joint distribution P (A; B), this method constructs a new variable T that extracts partitions, or clusters, over the values of A that are informative about B. The information bottleneck has already been applied to document classification, gene expression, neural code, and spectral analysis. In this paper, we introduce a general principled framework for multivariate extensions of the information bottleneck method. This allows us to consider multiple systems of data partitions that are inter-related. Our approach utilizes Bayesian networks for specifying the systems of clusters and what information each captures. We show that this construction provides insight about bottleneck variations and enables us to characterize solutions of these variations. We also present a general framework for iterative algorithms for constructing solutions, and apply it to several examples. IR snyders00confseek ConfSeek - A Multi-user, Multi-Threaded Specialized Search Engine for Conferences The explosive growth of the World Wide Web - the latest estimates for its size are around 1,000,000,000 web-pages - has made critical the need to nd information more accurately than what the current generic search engines can deliver. This project implements a prototype specialized search engine that allows user to submit queries for conferences in a specic eld of interest, and returns the detailed information about those conferences (deadlines, etc.). It uses multiple existing search engines to provide better coverage of the information available on the Internet. It can interact with multiple users concurrently and makes use of multi-threading to achieve faster information processing. The goal is to make this tool available to the scientic community to provide researchers with improved access to conference information. We discuss possible extensions (e.g. ranking of conferences according to their `quality', trip planning, etc.). Contents 1 Introduction 2 2 Goals of this Project... IR chang99automatic Automatic I/O Hint Generation through Speculative Execution Aggressive prefetching is an effective technique for reducing the execution times of disk-bound applications; that is, applications that manipulate data too large or too infrequently used to be found in file or disk caches. While automatic prefetching approaches based on static analysis or historical access patterns are effective for some workloads, they are not as effective as manually-driven (programmer-inserted) prefetching for applications with irregular or input-dependent access patterns. In this paper, we propose to exploit whatever processor cycles are left idle while an application is stalled on I/O by using these cycles to dynamically analyze the application and predict its future I/O accesses. Our approach is to speculatively pre-execute the application's code in order to discover and issue hints for its future read accesses. Coupled with an aggressive hint-driven prefetching system, this automatic approach could be applied to arbitrary applications, and should be particularl... DB 502499 Aliasing on the World Wide Web: Prevalence and Performance Implications Aliasing occurs in Web transactions when requests containing different URLs elicit replies containing identical data payloads. Aliasing can cause cache misses, and there is reason to suspect that offthe -shelf Web authoring tools might increase aliasing on the Web. Existing research literature, however, says little about the prevalence of aliasing in user-initiated transactions or its impact on endto -end performance in large multi-level cache hierarchies. IR 445103 Building Intelligent Systems For Mining Information Extraction Rules From Web Pages By Using Domain Knowledge Previous researches on automatic information extraction experienced difficulties in acquiring and representing useful domain knowledge and in coping with the structural heterogeneity among different information sources. As a result, many real-world information sources with complex document structures could not be correctly analyzed. In order to resolve these problems, this paper presents a method of building intelligent systems for mining information extraction rules from semi-structured Web pages by using domain knowledge. This system automatically generates a wrapper for each information source and performs information extraction and information integration by applying this wrapper to the corresponding source. Both the domain knowledge and the wrapper are represented by XML documents to increase flexibility and interoperability. By testing our prototype system on several real-estate information sites, we can claim that it creates the correct wrappers for most Web sources and consequently facilitates effective information extraction for heterogeneous information sources. 1. IR 453805 Using Models of Score Distributions in Information Retrieval Empirical modeling of a number of different text search engines shows that the score distributions on a per query basis may be fitted approximately using an exponential distribution for the set of nonrelevant documents and a normal distribution for the set of relevant documents. This model fits not only probabilistic search engines like INQUERY but also vector space search engines like SMART and also LSI search engines. The model also appears to be true of search engines operating on a number of different languages. This leads to the hypothesis that all 'good' text search engines operating on any language have similar characteristics. The question then arises as to whether the shape of the score distributions reflects some underlying model of language or the search process itself. We discuss how they arise given certain assumptions about word distributions in documents. We then show that given a query for which relevance information is not available, a mixture model consisting of an exponential and a normal distribution can be fitted to the score distribution. These distributions can be used to map the scores of a search engine to probabilities. This model has many possible applications. For example, the outputs of different search engines can be combined by averaging the probabilities (optimal if the search engines are independent) or by using the probabilities to select the best engine for each query. Results show that the technique performs as well as the best current combination techniques. A number of different IR tasks may benefit from score modeling including filtering, multi-lingual retrieval and relevance feedback. We also discuss possible future improvements to the process of score modeling. 1. IR norrie00extended An Extended Entity-Relationship Approach to Data Management in Object-Oriented Systems Database programming in object-oriented systems can be supported by combining data modelling and programming technologies such that a data model supports the management of collections of objects where those objects are as specified by the underlying object-oriented programming language. This approach is the basis of the object data management services (ODMS) of the Comandos system. The ODMS data model provides constructs for the representation of both entities and their relationships and further supports rich classification structures. To complement the structural model, there is an operational model based on an algebra over collections of objects. 1 Introduction Object-oriented technologies are gaining in popularity as the basis for software development platforms. Meanwhile the family of entity-relationship data models retain their wide-spread use and popularity for conceptual modelling. How then can these two successful technologies be combined to support the development of ... DB bohlin00path Path Planning Using Lazy PRM This paper describes a new approach to probabilistic roadmap planners (PRMs). The overall theme of the algorithm, called Lazy PRM, is to minimize the number of collision checks performed during planning and hence minimize the running time of the planner. Our algorithm builds a roadmap in the configuration space, whose nodes are the user-defined initial and goal configurations and a number of randomly generated nodes. Neighboring nodes are connected by edges representing paths between the nodes. In contrast with PRMs, our planner initially assumes that all nodes and edges in the roadmap are collision-free, and searches the roadmap at hand for a shortest path between the initial and the goal node. The nodes and edges along the path are then checked for collision. If a collision with the obstacles occurs, the corresponding nodes and edges are removed from the roadmap. Our planner either finds a new shortest path, or first updates the roadmap with new nodes and edges, and then searches for a shortest path. The above process is repeated until a collision-free path is returned. AI giorgini01multiagent Multi-Agent Architectures as Organizational Structures A Multi-Agent System (MAS) is an organization of coordinated autonomous agents that interact in order to achieve particular, possible common goals. Considering real world organizations as an analogy, this paper proposes architectural styles for MAS which adopt concepts from organizational theories. The styles are modeled using the i* framework which o#ers the notions of actor, goal and actor dependency and specified in Formal Tropos. They are evaluated with respect to a set of software quality attributes, such as predictability or adaptability. In addition, we conduct a comparative study of organizational and conventional software architectures using the mobile robot control example from the Software Engineering literature. The research is conducted in the context of Tropos, a comprehensive software system development methodology. Agents grieser00unifying A Unifying Approach to HTML Wrapper Representation and Learning . The number, the size, and the dynamics of Internet information sources bears abundant evidence of the need for automation in information extraction. This calls for representation formalisms that match the World Wide Web reality and for learning approaches and learnability results that apply to these formalisms. The concept of elementary formal systems is appropriately generalized to allow for the representation of wrapper classes which are relevant to the description of Internet sources in HTML format. Related learning results prove that those wrappers are automatically learnable from examples. This is setting the stage for information extraction from the Internet by exploitation of inductive learning techniques. 1 Motivation Today's online access to millions or even billions of documents in the World Wide Web is a great challenge to research areas related to knowledge discovery and information extraction (IE). The general task of IE is to locate specific pieces of text i... IR 455511 Probabilistic Default Reasoning with Conditional Constraints We present an approach to reasoning from statistical and subjective knowledge, which is based on a combination of probabilistic reasoning from conditional constraints with approaches to default reasoning from conditional knowledge bases. More precisely, we introduce the notions of -, lexicographic, and conditional entailment for conditional constraints, which are probabilistic generalizations of Pearl's entailment in system , Lehmann's lexicographic entailment, and Geffner's conditional entailment, respectively. We show that the new formalisms have nice properties. In particular, they show a similar behavior as referenceclass reasoning in a number of uncontroversial examples. The new formalisms, however, also avoid many drawbacks of reference-class reasoning. More precisely, they can handle complex scenarios and even purely probabilistic subjective knowledge as input. Moreover, conclusions are drawn in a global way from all the available knowledge as a whole. We then show that the new formalisms also have nice general nonmonotonic properties. In detail, the new notions of -, lexicographic, and conditional entailment have similar properties as their classical counterparts. In particular, they all satisfy the rationality postulates proposed by Kraus, Lehmann, and Magidor, and they have some general irrelevance and direct inference properties. Moreover, the new notions of - and lexicographic entailment satisfy the property of rational monotonicity. Furthermore, the new notions of -, lexicographic, and conditional entailment are proper generalizations of both their classical counterparts and the classical notion of logical entailment for conditional constraints. Finally, we provide algorithms for reasoning under the new formalisms, and we analyze its computational com... DB sakama99updating Updating Extended Logic Programs through Abduction . This paper introduces techniques for updating knowledge bases represented in extended logic programs. Three di#erent types of updates, view updates, theory updates, and inconsistency removal, are considered. We formulate these updates through abduction, and provide methods for computing them with update programs. An update program is an extended logic program which specifies changes on abductive hypotheses, then updates are computed by the U-minimal answer sets of an update program. The proposed technique provides a uniform framework for these di#erent types of updates, and each update is computed using existing procedures of logic programming. 1 Introduction A knowledge base must be updated when new information arrives. There are three cases in updating a knowledge base. The first one is that a knowledge base contains two di#erent kinds of knowledge --- variable knowledge and invariable knowledge. In this case, updates are permitted only on variable knowledge. Updates on the invari... DB shavlik98building Building Intelligent Agents for Web-Based Tasks: A Theory-Refinement Approach We present and evaluate an infrastructure with which to rapidly and easily build intelligent software agents for Web-based tasks. Our design is centered around two basic functions: ScoreThisLink and ScoreThisPage. If given highly accurate such functions, standard heuristic search would lead to efficient retrieval of useful information. Our approach allows users to tailor our system 's behavior by providing approximate advice about the above functions. This advice is mapped into neural network implementations of the two functions. Subsequent reinforcements from the Web (e.g., dead links) and any ratings of retrieved pages that the user wishes to provide are, respectively, used to refine the link- and page-scoring functions. Hence, our agent architecture provides an appealing middle ground between nonadaptive "agent" programming languages and systems that solely learn user preferences from the user's ratings of pages. We present a case study where we provide some simple advice and speci... ML koperski98mining Mining Knowledge in Geographical Data this article, a short overview is provided to summarize recent studies on spatial data mining, including spatial data mining techniques, their strengths and weaknesses, how and when to apply them, and what are the challenges yet to be faced. DB bernstein00vision A Vision for Management of Complex Models Many problems encountered when building applications of database systems involve the manipulation of models. By “model, ” we mean a complex structure that represents a design artifact, such as a relational schema, object-oriented interface, UML model, XML DTD, web-site schema, semantic network, complex document, or software configuration. Many uses of models involve managing changes in models and transformations of data from one model into another. These uses require an explicit representation of “mappings ” between models. We propose to make database systems easier to use for these applications by making “model ” and “model mapping ” first-class objects with special operations that simplify their use. We call this capability model management. In addition to making the case for model management, our main contribution is a sketch of a proposed data model. The data model consists of formal, object-oriented structures for representing models and model mappings, and of high-level algebraic operations on those structures, such as matching, differencing, merging, function application, selection, inversion and instantiation. We focus on structure and semantics, not implementation. 1 DB 509024 A Multidimentional Framework for the Evaluation of Multiagent System Methodologies Because of the great interest in using multiagent systems (MAS) in a wide variety of applications in recent years, agent-oriented methodologies and related modeling techniques have become a priority for the development of large scale agent-based systems. The work we present here belongs to the disciplines of Software Engineering and Distributed Artificial Intelligence. More specifically, we are interested in software engineering aspects involved in the development of multiagent systems (MAS). Several methodologies have been proposed for the development of MAS. For the most part, these methodologies remain incomplete: they are either an extension of object-oriented methodologies or an extension of knowledge-based methodologies. In addition, too little effort has gone into the standardization of MAS methodologies, platforms and environments. It seems obvious, therefore, that software engineering aspects of the development of MAS still remains an open field. The success of the agent paradigm requires systematic methodologies for the specification, analysis and design of "non toy" MAS applications. We present in this paper a new framework called MUCCMAS, which stands for MUltidimensional framework of Criteria for the Comparison of MAS methodologies, that enabled us to make a comparative analysis of existing main MAS methodologies. Agents funt98is Is Machine Colour Constancy Good Enough? . This paper presents a negative result: current machine colour constancy algorithms are not good enough for colour-based object recognition. This result has surprised us since we have previously used the better of these algorithms successfully to correct the colour balance of images for display. Colour balancing has been the typical application of colour constancy, rarely has it been actually put to use in a computer vision system, so our goal was to show how well the various methods would do on an obvious machine colour vision task, namely, object recognition. Although all the colour constancy methods we tested proved insufficient for the task, we consider this an important finding in itself. In addition we present results showing the correlation between colour constancy performance and object recognition performance, and as one might expect, the better the colour constancy the better the recognition rate. 1 Introduction We set out to show that machine colour constancy had matured to... ML itti98model A Model of Saliency-based Visual Attention for Rapid Scene Analysis A visual attention system, inspired by the behavior and the neuronal architecture of the early primate visual system, is presented. Multiscale image features are combined into a single topographical saliency map. A dynamical neural network then selects attended locations in order of decreasing saliency. The system breaks down the complex problem of scene understanding by rapidly selecting, in a computationally efficient manner, conspicuous locations to be analyzed in detail. Index terms: Visual attention, scene analysis, feature extraction, target detection, visual search. \Pi I. Introduction Primates have a remarkable ability to interpret complex scenes in real time, despite the limited speed of the neuronal hardware available for such tasks. Intermediate and higher visual processes appear to select a subset of the available sensory information before further processing [1], most likely to reduce the complexity of scene analysis [2]. This selection appears to be implemented in the ... ML 162997 A Runtime System for Interactive Web Services Interactive web services are increasingly replacing traditional static web pages. Producing web services seems to require a tremendous amount of laborious lowlevel coding due to the primitive nature of CGI programming. We present ideas for an improved runtime system for interactive web services built on top of CGI running on virtually every combination of browser and HTTP/CGI server. The runtime system has been implemented and used extensively in , a tool for producing interactive web services. Keywords: CGI, Interactive Web Service, Web Document Management, Runtime System, Session Model. 1 Introduction An interactive web service consists of a global shared state (typically a database) and a number of distinct sessions that each contain some local private state and a sequential, imperative action. A web client may invoke an individual thread of one of the given session kinds. The execution of this thread may interact with the client and inspect or modify the global state. One... DB 29745 Rigid and Articulated Motion Seen with an Uncalibrated Stereo Rig This paper establishes a link between uncalibrated stereo vision and the motion of rigid and articulated bodies. The variation in the projective reconstruction of a dynamic scene over time allows an uncalibrated stereo rig to be used as a faithful motion capturing device. We introduce an original theoretical framework -- projective kinematics -- which allows rigid and articulated motion to be represented within the transformation group of projective space. Corresponding projective velocities are defined in the tangent space. Most importantly, these projective motions inherit the Lie-group structure of the displacement group. These theoretical results lead immediately to nonmetric formulations of visual servoing, tracking, motion capturing and motion synthesis systems, that no longer require the metric geometry of a stereo camera or of the articulated body to be known. We report on such a nonmetric formulation of a visual servoing system and present simulated experimental results. 1 In... AI andersson99intelligent Intelligent Agents -- A New Technology for Future Distributed Sensor Systems? This master thesis deals with intelligent agents and the possibility to use the intelligent agent technology in future distributed sensor systems. The term future distributed sensor system refers to a system based on several sensors that will be developed within a period of five to ten years. Since researchers have not agreed on a more precise definition of intelligent agents, we first examined what constitutes an intelligent agent and made a definition suited for our application domain. We used our definition as a base for investigating if and how intelligent agents can be used in future distributed sensor systems. We argue that it is not interesting to come up with a general agent definition applicable to every agent, instead one should make a foundation for a definition. When this is done we can decide on more specific features depending on the task the agent will perform and in what domain the agent will work in. Finally we conclude that it is possible to use the agent technology i... Agents breazeal98regulating Regulating Human-Robot Interaction using "emotions", "drives" and facial expressions This paper presents a motivational system for an autonomous robot which is designed to regulate humanrobot interaction. The mode of social interaction is that of a caretaker-infant dyad where a human acts as the caretaker for the robot. The robot's motivational system is designed to generate an analogous interaction for a robot-human dyad as for an infantcaretaker dyad. An infant's emotions and drives play a very important role in generating meaningful interactions with the caretaker (Bullowa 1979). Similarly, the learning task for the robot is to apply various communication skills acquired during social exchanges to manipulate the caretaker such that its drives are satisfied. Toward this goal, the motivational system implements drives, emotions, and facial expressions. The interaction is regulated specifically to promote a suitable learning environment. Although the details of the learning itself are beyond the scope of this paper, this work represents an important step toward realiz... AI 376260 SQL Based Association Rule Mining using Commercial RDBMS (IBM DB2 UDB EEE) . Data mining is becoming increasingly important since the size of databases grows even larger and the need to explore hidden rules from the databases becomes widely recognized. Currently database systems are dominated by relational database and the ability to perform data mining using standard SQL queries will definitely ease implementation of data mining. However the performance of SQL based data mining is known to fall behind specialized implementation and expensive mining tools being on sale. In this paper we present an evaluation of SQL based data mining on commercial RDBMS (IBM DB2 UDB EEE). We examine some techniques to reduce I/O cost by using View and Subquery. Those queries can be more than 6 times faster than SETM SQL query reported previously. In addition, we have made performance evaluation on parallel database environment and compared the performance result with commercial data mining tool (IBM Intelligent Miner). We prove that SQL based data mining can achie... DB 205160 Analysis on a Mobile Agent Based Algorithm for Network Management Recent advance in the agent technology has brought in a new method for network routing, the ant routing algorithm. Although its effectiveness and efficiency have been demonstrated and reported in the literature, its properties have not yet been well studied. This paper will present some preliminary analysis of this algorithm in regard to its population growing property and jumping behavior. For both synchronized and asynchronized networks, we have shown that the expected number of agents in a node is no more than (1 +max i fj\Omega i jg)km, where j\Omega i j is the number of neighbor hosts of the i th host; k is the number of agents generated per request and m is the average number of requests. It is shown that under a mild condition, for all p (1 + max i fj\Omega i jg)km, the probability of the number of agents in a node exceeding p is less than R 1 p P(x)dx; where P(x) is a normal distributed function with mean and variance given by (1+max i fj\Omega i jg)km and (km)... Agents franconi00icom The i.com Tool for Intelligent Conceptual Modelling In this paper we present i.com, a tool for intelligent conceptual modelling. i.com allows for the specification of multiple EER diagrams and inter- and intra-schema constraints. Complete logical reasoning is employed by the tool to verify the specification, infer implicit facts, and manifest any inconsistencies. 1 Introduction i.com is a tool supporting the conceptual design phase of an information system, and in particular of an integration information system -- such as a data warehouse. The tool is an evolution of part of the conceptual modelling demonstrators suite [Jarke et al., 2000] developed within the European ESPRIT Long Term Research Data Warehouse Quality (DWQ) project [Jarke et al., 1999] . i.com adopts an extended Entity-Relationship (EER) conceptual data model, enriched with multidimensional aggregations and interschema constraints. i.com is fully integrated with a very powerful description logics reasoning server which acts as a background inference engine. The co... DB zhang99situated Situated Neuro-Fuzzy Control for Vision-Based Robot Localisation We introduce a neuro-fuzzy system for localising mobile robot solely based on raw vision data without relying on landmarks or artificial symbols. In an initial learning step the system is trained on the compressed input data so as to classify different situations and to associate appropriate behaviours to these situations. Input data may, for example, be generated by an omnidirectional vision system obviating the need for active cameras. At run time the compressed input data are fed into different B-spline fuzzy controllers which determine the correspondence between the actual situation and the situation they were trained for. The matching controller may then directly drive the actuators to realise the desired behaviour. The system thus realises a tight coupling between a very high-dimensional input parameter space and the robot actuators. It is completely free of any internal models such as maps of the environment, the algorithms are straightforward to implement and the computational ... ML zhang01efficient Efficient Computation of Temporal Aggregates with Range Predicates A temporal aggregation query is an important but costly operation for applications that maintain timeevolving data (data warehouses, temporal databases, etc.). Due to the large volume of such data, performance improvements for temporal aggregation queries are critical. In this paper we examine techniques to compute temporal aggregates that include key-range predicates (range temporal aggregates). In particular we concentrate on SUM, COUNT and AVG aggregates. This problem is novel; to handle arbitrary key ranges, previous methods would need to keep a separate index for every possible key range. We propose an approach based on a new index structure called the Multiversion SB-Tree, which incorporates features from both the SB-Tree and the Multiversion B-Tree, to handle arbitrary key-range temporal SUM, COUNT and AVG queries. We analyze the performance of our approach and present experimental results that show its efficiency. 1 DB holmquist00play The PLAY Research Group: Entertainment and Innovation in Sweden In a short time the research group PLAY has established an unorthodox but effective work style, where a creative approach to research in information technology is combined with a strong focus on achieving high-quality results. Being a young research group (both regarding the time it has existed and the average age of its members) has presented PLAY with both challenges and opportunities. We face the challenge of building a credible basis for research in the academic community, but also think that we have the opportunity to contribute innovative results to the research community and our industrial partners. Keywords HCI research groups, future HCI, European HCI, IT design INTRODUCTION How can one perform exciting and unorthodox research in information technology, while still assuring that results are useful and of good quality? How can a small group, consisting mostly of relatively inexperienced students, in a small country with very little traditions in groundbreaking IT research, ... HCI cadoli98survey A Survey on Knowledge Compilation this paper we survey recent results in knowledge compilation of propositional knowledge bases. We first define and limit the scope of such a technique, then we survey exact and approximate knowledge compilation methods. We include a discussion of compilation for non-monotonic knowledge bases. Keywords: Knowledge Representation, Efficiency of Reasoning AI silverman01more More Realistic Human Behavior Models for Agents in Virtual Worlds: Emotion, Stress, and Value Ontologies This paper focuses on challenges to improving the behavioral realism of computer generated agents and attempts to reflect the state of the art in human behavior modeling with particular attention to value ontologies, emotion, and stress in game-theoretic settings. The goal is to help those interested in constructing more realistic software agents for use in simulations, in virtual reality environments, and in training and performance aiding settings such as on the web or in embedded applications. This paper pursues this goal by providing a framework for better integrating the theories and models contained in the diverse human behavior modeling literatures, such as those that straddle physiological, cognitive and emotive processes; individual differences; emergent group and crowd behavior; and (punctuated) equilibria in social settings. The framework is based on widely available ontologies of world values and how these and physiological factors might be construed emotively into subjective expected utilities to guide the reactions and deliberations of agents. For example what makes one set of opponent groups differ from another? This framework serves as an extension of Markov decision processes appropriate for iterative play in game-theoretic settings, with particular emphasis on agent capabilities for redefining drama and for finding meta-games to counter the human player. This article presents the derivation of the framework and some initial results and lessons learned about integrating behavioral models into interactive dramas and meta-games that stimulate (systemic) thought and training doctrine. 1) Agents bohlen00temporal Temporal Statement Modifiers A wide range of database applications manage time-varying data. Many temporal query languages have been proposed, each one the result of many carefully made yet subtly interacting design decisions. In this article we advocate a different approach to articulating a set of requirements, or desiderata, that directly imply the syntactic structure and core semantics of a temporal extension of an (arbitrary) nontemporal query language. These desiderata facilitate transitioning applications from a nontemporal query language and data model, which has received only scant attention thus far. The paper then introduces the notion of statement modifiers that provide a means of systematically adding temporal support to an existing query language. Statement modifiers apply to all query language statements, for example, queries, cursor definitions, integrity constraints, assertions, views, and data manipulation statements. We also provide a way to systematically add temporal support to an existing implementation. The result is a temporal query language syntax, semantics, and implementation that derives from first principles. We exemplify this approach by extending SQL-92 with statement modifiers. This extended language, termed ATSQL, is formally defined via a denotational-semantics-style mapping of DB ezeife01selecting Selecting and Materializing Horizontally Partitioned Warehouse Views Data warehouse views typically store large aggregate tables based on a subset of dimension attributes of the main data warehouse fact table. Aggregate views can be stored as 2 n subviews of a data cube with n attributes. Methods have been proposed for selecting only some of the data cube views to materialize in order to speed up query response time, accommodate storage space constraint and reduce warehouse maintenance cost. This paper proposes a method for selecting and materializing views, which selects and horizontally fragments a view, recomputes the size of the stored partitioned view while deciding further views to select. # 2001 Elsevier Science B.V. All rights reserved. Keywords: Data warehouse; Views; Fragmentation; Performance benet 1. Introduction Decision support systems (DSS) used by business executives require analyzing snapshots of departmental databases over several periods of time. Departmental databases of the same organization (e.g., a bank) may be stored on dier... DB parker98adaptive Adaptive Heterogeneous Multi-Robot Teams This research addresses the problem of achieving fault tolerant cooperation within small- to medium-sized teams of heterogeneous mobile robots. We describe a novel behavior-based, fully distributed architecture, called ALLIANCE, that utilizes adaptive action selection to achieve fault tolerant cooperative control in robot missions involving loosely coupled tasks. The robots in this architecture possess a variety of high-level functions that they can perform during a mission, and must at all times select an appropriate action based on the requirements of the mission, the activities of other robots, the current environmental conditions, and their own internal states. Since such cooperative teams often work in dynamic and unpredictable environments, the software architecture allows the team members to respond robustly and reliably to unexpected environmental changes and modi cations in the robot team that may occur due to mechanical failure, the learning of new skills, or the addition or removal of robots from the team by human intervention. After presenting ALLIANCE, we describe in detail our experimental results of an implementation of this architecture on a team of physical mobile robots performing a cooperative box pushing demonstration. These experiments illustrate the ability ofALLIANCE to achieve adaptive, fault-tolerant cooperative control amidst dynamic changes in the capabilities of the robot team. AI critchlow00datafoundry DataFoundry: Information Management for Scientific Data Data warehouses and data marts have been successfully applied to a multitude of commercial business applications. They have proven to be invaluable tools by integrating information from distributed, heterogeneous sources and summarizing this data for use throughout the enterprise. Although the need for information dissemination is as vital in science as in business, working warehouses in this community are scarce because traditional warehousing techniques don't transfer to scientific environments. There are two primary reasons for this difficulty. First, schema integration is more difficult for scientific databases than for business sources, because of the complexity of the concepts and the associated relationships. While this difference has not yet been fully explored, it is an important consideration when determining how to integrate autonomous sources. Second, scientific data sources have highly dynamic data representations (schemata). When a data source participating in a warehouse... DB liu99deductive Deductive Database Languages: Problems and Solutions this paper, we discuss these problems from four different aspects: complex values, object orientation, higher-orderness, and updates. In each case, we examine four typical languages that address the corresponding issues. DB herbrich99regression Regression Models for Ordinal Data: A Machine Learning Approach In contrast to the standard machine learning tasks of classification and metric regression we investigate the problem of predicting variables of ordinal scale, a setting referred to as ordinal regression. The task of ordinal regression arises frequently in the social sciences and in information retrieval where human preferences play a major role. Also many multi--class problems are really problems of ordinal regression due to an ordering of the classes. Although the problem is rather novel to the Machine Learning Community it has been widely considered in Statistics before. All the statistical methods rely on a probability model of a latent (unobserved) variable and on the condition of stochastic ordering. In this paper we develop a distribution independent formulation of the problem and give uniform bounds for our risk functional. The main difference to classification is the restriction that the mapping of objects to ranks must be transitive and asymmetric. Combining our theoretical framework with results from measurement theory we present an approach that is based on a mapping from objects to scalar utility values and thus guarantees transitivity and asymmetry. Applying the principle of Structural Risk Minimization as employed in Support Vector Machines we derive a new learning algorithm based on large margin rank boundaries for the task of ordinal regression. Our method is easily extended to nonlinear utility functions. We give experimental results for an Information Retrieval task of learning the order of documents with respect to an initial query. Moreover, we show that our algorithm outperforms more naive approaches to ordinal regression such as Support Vector Classification and Support Vector Regression in the case of more than two ranks. IR druin01designing Designing a Digital Library for Young Children: An Intergenerational Partnership As more information resources become accessible using computers, our digital interfaces to those resources need to be appropriate for all people. However when it comes to digital libraries, the interfaces have typically been designed for older children or adults. Therefore, we have begun to develop a digital library interface developmentally appropriate for young children (ages 5-10 years old). Our prototype system we now call "QueryKids" offers a graphical interface for querying, browsing and reviewing search results. This paper describes our motivation for the research, the design partnership we established between children and adults, our design process, the technology outcomes of our current work, and the lessons we have learned. Keywords Children, digital libraries, information retrieval design techniques, education applications, participatory design, cooperative inquiry, intergenerational design team, zoomable user interfaces (ZUIs). THE NEED FOR RESEARCH A growing body of k... IR 337032 A Sound Algorithm for Region-Based Image Retrieval Using an Index Region-based image retrieval systems aim to improve the effectiveness of content-based search by decomposing each image into a set of "homogeneous" regions. Thus, similarity between images is assessed by computing similarity between pairs of regions and then combining the results at the image level. In this paper we propose the first provably sound algorithm for performing region-based similarity search when regions are accessed through an index. Experimental results demonstrate the effectiveness of our approach, as also compared to alternative retrieval strategies. 1. Introduction Many real world applications, in the field of medicine, weather prediction, and communications, to name a few, require efficient access to image databases based on content. To this end, the goal of content-based image retrieval (CBIR) systems is to define a set of properties (features) able to effectively characterize the content of images and then to use such features during retrieval. Users accessing a CB... ML 150449 A Wearable Spatial Conferencing Space Wearable computers provide constant access to computing and communications resources. In this paper we describe how the computing power of wearables can be used to provide spatialized 3D graphics and audio cues to aid communication. The result is a wearable augmented reality communication space with audio enabled avatars of the remote collaborators surrounding the user. The user can use natural head motions to attend to the remote collaborators, can communicate freely while being aware of other side conversations and can move through the communication space. In this way the conferencing space can support dozens of simultaneous users. Informal user studies suggest that wearable communication spaces may offer several advantages, both through the increase in the amount of information it is possible to access and the naturalness of the interface. 1: Introduction One of the broad trends emerging in human-computer interaction is the increasing portability of computing and communication fac... HCI 89078 Confluence of Computer Vision and Interactive Graphics for Augmented Reality . Augmented reality #AR# is a technology in which a user's view of the real world is enhanced or augmented with additional information generated from a computer model. Using AR technology, users can interact with a combination of real and virtual objects in a natural way. This paradigm constitutes the core of a very promising new technology for many applications. However, before it can be applied successfully, AR has to ful#ll very strong requirements including precise calibration, registration and tracking of sensors and objects in the scene, as well as a detailed overall understanding of the scene. At ECRCwe see computer vision and image processing technology play an increasing role in acquiring appropriate sensor and scene models. To balance robustness with automation, weintegrate automatic image analysis with both interactive user assistance and input from magnetic trackers and CAD-models. Also, in order to meet the requirements of the emerging global information society... HCI carreras01hybrid Hybrid Coordination of Reinforcement Learning-based Behaviors for AUV Control This paper proposes a Hybrid Coordination method for Behavior-based Control Architectures. The hybrid method takes in advantages of the robustness and modularity in competitive approaches as well as optimized trajectories in cooperative ones. This paper will demonstrate the feasibility of this hybrid method with a 3D-navigation application to an Autonomous Underwater Vehicle (AUV). The behaviors were learnt online by means of Reinforcement Learning. Q(l)- learning was used extending the one-step learning of the popular Q-learning to n-steps. Realistic simulations were carried out. Results showed the good performance of the hybrid method on behavior coordination as well as on increasing and improving behavior learning. ML 87928 Authoritative Sources in a Hyperlinked Environment The link structure of a hypermedia environment can be a rich source of information about the content of the environment, provided we have effective means for understanding it. Versions of this principle have been studied in the hypertext research community and (in a context predating hypermedia) through journal citation analysis in the field of bibliometrics. But for the problem of searching in hyperlinked environments such as the World Wide Web, it is clear from the prevalent techniques that the information inherent in the links has yet to be fully exploited. In this work we develop a new method for automatically extracting certain types of information about a hypermedia environment from its link structure, and we report on experiments that demonstrate its effectiveness for a variety of search problems on the www. The central problem we consider is that of determining the relative "authority" of pages in such environments. This issue is central to a number of basic hypertext search t... IR 267537 Hybrid Neural Systems This chapter provides an introduction to the eld of hybrid neural systems. Hybrid neural systems are computational systems which are based mainly on articial neural networks but also allow a symbolic interpretation or interaction with symbolic components. In this overview, we will describe recent results of hybrid neural systems. We will give a brief overview of the main methods used, outline the work that is presented here, and provide additional references. We will also highlight some important general issues and trends. ML 186486 Design Issues for Mixed-Initiative Agent Systems This paper addresses the effect of mixed-initiative systems on multiagent systems design. A mixed-initiative system is one in which humans interact directly with software agents in a collaborative approach to problem solving. There are two main levels at which multiagent systems are designed: the domain level and the individual agent level. At the domain level, there are few unique challenges to mixedinitiative system design. However, at the individual agent level, the agent itself must be designed to interact with the human and the agent system, integrating the two into a single system. Introduction Much of the current research related to intelligent agents has focused on the capabilities and structure of individual agents. However, in order to solve complex problems, these agents must work cooperatively with other agents in a heterogeneous environment. This is the domain of Multiagent Systems. In multiagent systems, we are interested in the coordinated behavior of a system of indiv... Agents bettini99symbolic Symbolic Representation of User-defined Time Granularities In the recent literature on time representation, an effort has been made to characterize the notion of time granularity and the relationships between granularities, in order to have a common framework for their specification, and to allow the interoperability of systems adopting different time granularities. This paper considers the mathematical characterization of finite and periodical time granularities, and it identifies a user-friendly symbolic formalism which captures exactly that class of granularities. This is achieved by a formal analysis of the expressiveness of well-known symbolic representation formalisms. 1. Introduction There is a wide agreement in the AI and database community on the requirement for a data/knowledge representation system of supporting standard as well as user-defined time granularities. Examples of standard time granularities are days, weeks, months, while user defined granularities may include businessweeks, trading-days, working-shifts, school-terms, wi... DB 147521 Query Optimization in the Presence of Limited Access Patterns 1 Introduction The goal of a query optimizer of a database system is to translate a declarative query expressed on a logical schema into an imperative query execution plan that accesses the physical storage of the data, and applies a sequence of relational operators. In building query execution plans, traditional relational query optimizers try to find the most efficient method for accessing the necessary data. When possible, a query optimizer will use auxiliary data structures such as an index on a file in order to efficiently retrieve a certain set of tuples in a relation. However, when such structures do not exist or are not useful for the given query, the alternative of scanning the entire relation always exists. The existence of the fall back option to perform a complete scan is an important assumption in traditional query optimization. Several recent query processing applications have the common characteristic that it is not always possible to perform complete scans on the data. Instead, the query optimization problem is complicated by the fact that there are only limited access patterns to the data. One such DB namee01proposal A Proposal for an Agent Architecture for Proactive Persistent Non Player Characters In the past, games development has been driven by the need to achieve more realistic graphics. Agents lawrence99text Text and Image Metasearch on the Web As the Web continues to increase in size, the relative coverage of Web search engines is decreasing, and search tools that combine the results of multiple search engines are becoming more valuable. This paper provides details of the text and image metasearch functions of the Inquirus search engine developed at the NEC Research Institute. For text metasearch, we describe features including the use of link information in metasearch, and provide statistics on the usage and performance of Inquirus and the Web search engines. For image metasearch, Inquirus queries multiple image search engines on the Web, downloads the actual images, and creates image thumbnails for display to the user. Inquirus handles image search engines that return direct links to images, and engines that return links to HTML pages. For the engines that return HTML pages, Inquirus analyzes the text on the pages in order to predict which images are most likely to correspond to the query. The individual image search engin... IR agrawal00athena Athena: Mining-based Interactive Management of Text Databases Abstract. We describe Athena: a system for creating, exploiting, and maintaining a hierarchy of textual documents through interactive miningbased operations. Requirements of any such system include speed and minimal end-user e ort. Athena satis es these requirements through linear-time classi cation and clustering engines which are applied interactively to speed the development of accurate models. Naive Bayes classi ers are recognized to be among the best for classifying text. We show that our specialization of the Naive Bayes classi er is considerably more accurate (7 to 29 % absolute increase in accuracy) than a standard implementation. Our enhancements include using Lidstone's law of succession instead of Laplace's law, under-weighting long documents, and over-weighting author and subject. We also present a new interactive clustering algorithm, C-Evolve, for topic discovery. C-Evolve rst nds highly accurate cluster digests (partial clusters), gets user feedback to merge and correct these digests, and then uses the classi cation algorithm to complete the partitioning of the data. By allowing this interactivity in the clustering process, C-Evolve achieves considerably higher clustering accuracy (10 to 20 % absolute increase in our experiments) than the popular K-Means and agglomerative clustering methods. 1 IR muhlenbein99convergence Convergence Theory and Applications of the Factorized Distribution Algorithm The paper investigates the optimization of additively decomposable functions (ADF) by a new evolutionary algorithm called Factorized Distribution Algorithm (FDA). FDA is based on a factorization of the distribution to generate search points. First separable ADFs are considered. These are mapped to generalized linear functions with metavariables defined for multiple alleles. The mapping transforms FDA into an Univariate Marginal Frequency Algorithm (UMDA). For UMDA the exact equation for the response to selection is computed under the assumption of proportionate selection. For truncation selection an approximate equation for the time to convergence is used, derived from an analysis of the OneMax function. FDA is also numerically investigated for non separable functions. The time to convergence is very similar to separable ADFs. FDA outperforms the genetic algorithm with recombination of strings by far. Keywords response to selection, Fisher's Theorem, additively decomposable functions... ML 251052 Control Law Design for Haptic Interfaces to Virtual Reality The goal of control law design for haptic displays is to provide a safe and stable user interface while maximizing the operator's sense of kinesthetic immersion in a virtual environment. This paper outlines a control design approach which stabilizes a haptic interface when coupled to a broad class of human operators and virtual environments. Two-port absolute stability criteria are used to develop explicit control law design bounds for two different haptic display implementations: impedance display and admittance display. The strengths and weaknesses of each approach are illustrated through numerical and experimental results for a three degree-offreedom device. The example highlights the ability of the proposed design procedure to handle some of the more difficult problems in control law synthesis for haptics, including structural flexibility and non-collocation of sensors and actuators. The authors are with the Department of Electrical Engineering University of Washington, Box 352500 Seattle, WA 98195-2500 2 I. HCI laviola99flex Flex And Pinch: A Case Study Of Whole Hand Input Design For Virtual Environment Interaction We present a discussion of design issues involving whole hand input in virtual environments. In many cases, whole hand input devices limit the types of interaction that the user can perform in the virtual world due to the nature of the device. One possible approach to alleviate these limitations is to provide hybrid input devices which enable the user to combine information generated from two different whole hand input devices. In this paper, we describe our Pinch Glove like input device which is used as a tool to augment bend-sensing gloves for object manipulation and menu selection as well as a method to test and evaluate different hand postures and gestures that could not be developed with a single whole hand device. KEYWORDS: Human-Computer Interaction, Virtual Environments, 3D Graphics Applications, Conductive Cloth, Flex and Pinch Input INTRODUCTION There have been a number of different approaches for interacting in virtual environments. In general, these approaches have attem... HCI deligiannidis02dlove DLoVe: Using Constraints to Allow Parallel Processing in Multi-User Virtual Reality In this paper, we introduce DLoVe, a new paradigm for designing and implementing distributed and nondistributed virtual reality applications, using one-way constraints. DLoVe allows programs written in its framework to be executed on multiple computers for improved performance. It also allows easy specification and implementation of multi-user interfaces. DLoVe hides all the networking aspects of message passing among the machines in the distributed environment and performs the needed network optimizations. As a result, a user of DLoVe does not need to understand parallel and distributed programming to use the system; he or she needs only be able to use the serial version of the user interface description language. Parallelizing the computation is performed by DLoVe, without modifying the interface description. HCI casati01improving Improving Business Process Quality through Exception Understanding, Prediction, and Prevention Business process automation technologies are being increasingly used by many companies to improve the efficiency of both internal processes as well as of e-services offered to customers. In order to satisfy customers and employees, business processes need to be executed with a high and predictable quality. In particular, it is crucial for organizations to meet the Service Level Agreements (SLAs) stipulated with the customers and to foresee as early as possible the risk of missing SLAs, in order to set the right expectations and to allow for corrective actions. In this paper we focus on a critical issue in business process quality: that of analyzing, predicting and preventing the occurrence of exceptions, i.e., of deviations from the desired or acceptable behavior. We characterize the problem and propose a solution, based on data warehousing and mining techniques. We then describe the architecture and implementation of a tool suite that enables exception analysis, prediction, and prevention. Finally, we show experimental results obtained by using the tool suite to analyze internal HP processes. 1. DB panzarasa02formalizing Formalizing Collaborative Decision-making and Practical Reasoning in Multi-agent Systems In this paper, we present an abstract formal model of decision-making in a social setting that covers all aspects of the process, from recognition of a potential for cooperation through to joint decision. In a multi-agent environment, where self-motivated autonomous agents try to pursue their own goals, a joint decision cannot be taken for granted. In order to decide effectively, agents need the ability to (a) represent and maintain a model of their own mental attitudes, (b) reason about other agents' mental attitudes, and (c) influence other agents' mental states. Social mental shaping is advocated as a general mechanism for attempting to have an impact on agents' mental states in order to increase their cooperativeness towards a joint decision. Our approach is to specify a novel, high-level architecture for collaborative decision-making in which the mentalistic notions of belief, desire, goal, intention, preference and commitment play a central role in guiding the individual agent's and the group's decision-making behaviour. We identify preconditions that must be fulfilled before collaborative decision-making can commence and prescribe how cooperating agents should behave, in terms of their own decision-making apparatus and their interactions with others, when the decision-making process is progressing satisfactorily. The model is formalized through a new, many-sorted, multi-modal logic. Agents heinze99plan Plan Recognition in Military Simulation: Incorporating Machine Learning with Intelligent Agents A view of plan recognition shaped by both operational and computational requirements is presented. Operational requirements governing the level of fidelity and nature of the reasoning process combine with computational requirements including performance speed and software engineering effort to constrain the types of solutions available to the software developer. By adopting machine learning to provide spatio-temporal recognition of environmental events and relationships, an agent can be provided with a mechanism for mental state recognition qualitatively different from previous research. An architecture for integrating machine learning into a BDI agent is suggested and the results from the development of a prototype provide proof-of-concept. 1 Introduction This paper proposes machine-learning as a tool to assist in the construction of agents capable of plan recognition. This paper focuses on the beliefs-desires-intentions (BDI) class of agents. These agents have been ... Agents paulson00inductive The Inductive Approach to Verifying Cryptographic Protocols Informal arguments that cryptographic protocols are secure can be made rigorous using inductive definitions. The approach is based on ordinary predicate calculus and copes with infinite-state systems. Proofs are generated using Isabelle/HOL. The human e#ort required to analyze a protocol can be as little as a week or two, yielding a proof script that takes a few minutes to run. Protocols are inductively defined as sets of traces. A trace is a list of communication events, perhaps comprising many interleaved protocol runs. Protocol descriptions incorporate attacks and accidental losses. The model spy knows some private keys and can forge messages using components decrypted from previous tra#c. Three protocols are analyzed below: OtwayRees (which uses shared-key encryption), Needham-Schroeder (which uses public-key encryption), and a recursive protocol [9] (which is of variable length). One can prove that event ev always precedes event ev # or that property P holds provided X remains ... ML slonim01power The Power of Word Clusters for Text Classification The recently introduced Information Bottleneck method [21] provides an information theoretic framework, for extracting features of one variable, that are relevant for the values of another variable. Several previous works already suggested applying this method for document clustering, gene expression data analysis, spectral analysis and more. In this work we present a novel implementation of this method for supervised text classification. Specifically, we apply the information bottleneck method to find word-clusters that preserve the information about document categories and use these clusters as features for classification. Previous work [1] used a similar clustering procedure to show that word-clusters can significantly reduce the feature space dimensionality, with only a minor change in classification accuracy. In this work we present similar results and go further to show that when the training sample is small word clusters can yield significant improvement in classification accuracy (up to  ¢¡¤£) over the performance using the words directly. 1 IR lawrence98searching Searching the world wide web The coverage and recency of the major World Wide Web search engines was analyzed, yielding some surprising results. The coverage of any one engine is significantly limited: No single engine indexes more than about one-third of the “indexable Web, ” the coverage of the six engines investigated varies by an order of magnitude, and combining the results of the six engines yields about 3.5 times as many documents on average as compared with the results from only one engine. Analysis of the overlap between pairs of engines gives an estimated lower bound on the size of the indexable Web of 320 million pages. The Internet has grown rapidly since its inception in December 1969 (1) and is anticipated to expand 1000 % over the next few years (2). The amount of scientific information and the number of electronic journals on the Internet continue to increase [about 1000 journals as of 1996 (2, 3)]. The Internet and the World Wide Web (the Web) represent significant advancements for the retrieval and dissemination of scientific and other literature and for the advancement of education (2, 4). With the introduction of full-text search engines such as AltaVista (www. IR 305534 An Open Framework for Distributed Multimedia Retrieval This article describes a framework for distributed multimedia retrieval which permits the connection of compliant user interfaces with a variety of multimedia retrieval engines via an open communication protocol, MRML (Multi Media Retrieval Markup Language). It allows the choice of image collection, feature set and query algorithm during run{time, permitting multiple users to query a system adapted to their needs, using the query paradigm adapted to their problem such as query by example (QBE), browsing queries, or query by annotation. User interaction is implemented over several levels and in diverse ways. Relevance feedback is implemented using positive and negative example images that can be used for a best{match QBE query. In contrast, browsing methods try to approach the searched image by giving overviews of the entire collection and by successive renement. In addition to these query methods, Long term o line learning is implemented. It allows feature preferences per ... ML 441373 DB-Prism: Integrated Data Warehouses and Knowledge Networks for Bank Controlling DB-Prism is an integrated data warehouse system DB norman97motivationbased Motivation-Based Direction of Planning Attention in Agents With Goal Autonomy The action of an agent with goal autonomy will be driven by goals generated with reference to its own beliefs and desires. This ability is essential for agents that are required to act in their own interests in a domain that is not entirely predictable. At any time, the situation may warrant the generation of new goals. However, it is not always the case that changes in the domain that lead to the generation of a goal are detected immediately before the goal should be pursued. Action may not be appropriate for some time. Furthermore, an agent may be influenced by goals that tend to recur periodically, or at particular times of the day or week for example. Such goals serve to motivate an agent towards interacting with other agents or processes with certain types of predictable behaviour patterns. This thesis provides a model of a goal autonomous agent that may generate goals in response to unexpected changes in its domain or cyclically through automatic processes. An important effect of... Agents lin01efficiently Efficiently Computing Weighted Proximity Relationships in Spatial Databases Spatial data mining recently emerges from a number of real applications, such as real-estate marketing, urban planning, weather forecasting, medical image analysis, road traffic accident analysis, etc. It demands for efficient solutions for many new, expensive, and complicated problems. In this paper, we investigate the problem of evaluating the top k distinguished "features" for a "cluster" based on weighted proximity relationships between the cluster and features. We measure proximity in an average fashion to address possible nonuniform data distribution in a cluster. Combining a standard multi-step paradigm with new lower and upper proximity bounds, we presented an efficient algorithm to solve the problem. The algorithm is implemented in several different modes. Our experiment results not only give a comparison among them but also illustrate the efficiency of the algorithm. DB camacho00travelplan TravelPlan: A MultiAgent System to Solve Web Electronic Travel Problems This paper presents TravelPlan, a multiagent architecture to co-operative work between different elements (human and/or software) whose main goal is to recommend useful solutions in the electronic tourism domain to system users. The system uses different types of intelligent autonomous agents whose main characteristics are cooperation, negotiation, learning, planning and knowledge sharing. The information used by the intelligent agents is heterogeneous and geographically distributed. The main information source of the system is Internet (the web). Other information sources are distributed knowledge bases in the own system.. The process to obtain, filter, and store the information is performed automatically by agents. This information is translated into a homogeneous format for high-level reasoning in order to obtain different partial solutions. Partial solutions are reconstructed into a general solution (or solutions) to be presented to the user. The system will recommend different solution... Agents 4664 Computation of the Semantics of Autoepistemic Belief Theories Recently, one of the authors introduced a simple and yet powerful non-monotonic knowledge representation framework, called the Autoepistemic Logic of Beliefs, AEB. Theories in AEB are called autoepistemic belief theories. Every belief theory T has been shown to have the least static expansion T which is computed by iterating a natural monotonic belief closure operator \Psi T starting from T . This way, the least static expansion T of any belief theory provides its natural non-monotonic semantics which is called the static semantics. It is easy to see that if a belief theory T is finite then the construction of its least static expansion T stops after countably many iterations. However, a somewhat surprising result obtained in this paper shows that the least static expansion of any finite belief theory T is in fact obtained by means of a single iteration of the belief closure operator \Psi T (although this requires T to be of a special form, we also show that T can be always put in th... AI 9105 Partitioning-Based Clustering for Web Document Categorization Clustering techniques have been used by manyintelligent software agents in order to retrieve, lter, and categorize documents available on the World Wide Web. Clustering is also useful in extracting salient features of related web documents to automatically formulate queries and search for other similar documents on the Web. Traditional clustering algorithms either use a priori knowledge of document structures to de ne a distance or similarity among these documents, or use probabilistic techniques such as Bayesian classi cation. Many of these traditional algorithms, however, falter when the dimensionality of the feature space becomes high relative to the size of the document space. In this paper, we introduce two new clustering algorithms that can e ectively cluster documents, even in the presence of a very high dimensional feature space. These clustering techniques, which are based on generalizations of graph partitioning, do not require pre-speci ed ad hoc distance functions, and are capable of automatically discovering document similarities or associations. We conduct several experiments on real Web data using various feature selection heuristics, and compare our clustering schemes to standard distance-based techniques, such ashierarchical agglomeration clustering, and Bayesian classi cation methods, such as AutoClass. IR 318212 Towards Active Logic Programming In this paper we present the new logic programming language DALI, aimed at defining agents and agent systems. A main design objective for DALI has been that of introducing in a declarative fashion all the essential features, while keeping the language as close as possible to the syntax and semantics of the plain Horn--clause language. Special atoms and rules have been introduced, for representing: external events, to which the agent is able to respond (reactivity); actions (reactivity and proactivity); internal events (previous conclusions which can trigger further activity); past and present events (to be aware of what has happened). An extended resolution is provided, so that a DALI agent is able to answer queries like in the plain Horn--clause language, but is also able to cope with the different kinds of events, and exhibit a (rational) reactive and proactive behaviour. 1 Introduction In this paper we address the issue of defining a logic programming language for reac... Agents bakker01lart LART: flexible, low-power building blocks for wearable computers To ease the implementation of different wearable computers, we developed a low-power processor board (named LART) with a rich set of interfaces. The LART supports dynamic voltage scaling, so performance (and power consumption) can be scaled to match demands: 59-221 MHz, 106-640 mW. High-end wearables can be configured from multiple LARTs operating in parallel; alternatively, FPGA boards can be used for dedicated data-processing, which reduces power consumption significantly. HCI 16755 Case-based Learning for Knowledge-based Design Support . We present a general approach to combine methods of interactive knowledge acquisition with methods for machine learning. The approach has been developed in order to deliver knowledge required by support-systems for design-tasks. Learning rests upon a knowledge representation scheme for cases that distinguishes between knowledge needed for subgoaling and knowledge needed for design. We employ traces, i.e., protocols of the user's actions when tackling design-tasks as the initial input for incremental knowledge acquisition. This allows to learn task structures to be used for subgoaling and case-bases plus similarity relations applicable to particular case-bases. 1 INTRODUCTION Integrating incremental learning into a knowledge-based systems seems to be a promising way to lessen the burden of knowledge elicitation to system development [9]. The goal of this paper is to point out how learning can be used in an interactive design-support system that uses Cbr [8] as the main problem solvin... ML grasso99augmenting Augmenting Recommender Systems by Embedding Interfaces into Office Practices Automated collaborative filtering systems collect evaluations from users of the quality and relevance of stored information items, such as scientific papers, books, and movies. A number of users need to give evaluations for the systems to be able to produce statistically high quality predictions of an item's interest. Promoting the creation of a rich meta-layer of evaluations is essential for these systems, but several important issues remain to be resolved. The work presented here first analyses the issues around the collection of recommendations, then proposes a set of design principles for improving and automating the collection of recommendations, and finally presents how these principles have been implemented in a real usage setting. 1. Systems to alleviate the information overload `Information overload' may be an abused term but it is an increasingly apt description of our current experience in dealing with information. The increase in communication channels and publishing me... IR 508816 Enabling Ad-Hoc Collaboration through Schedule Learning and Prediction The transferal of the desktop interface to the world at large is not the goal of ubiquitous computing. Rather, ubiquitous computing strives to increase the responsiveness of the world at large to the individual. A large part of this responsiveness is improved communication with other individuals. In this paper we describe a system that can enable ad--hoc collaboration between several people by creating a model of the daily schedules of individuals and by performing predictions based on this model. Using GPS data we learn to distinguish locations and track the times that these locations are visited. In addition, we use Markov models to predict which locations might be visited next based on the user's previous behavior. HCI kanungo99omnipage OmniPage vs. Sakhr: Paired Model Evaluation of Two Arabic OCR Products Characterizing the performance of Optical Character Recognition (OCR) systems is crucial for monitoring technical progress, predicting OCR performance, providing scientific explanations for the system behavior and identifying open problems. While research has been done in the past to compare performances of two or more OCR systems, all assume that the accuracies achieved on individual documents in a dataset are independent when, in fact, they are not. In this paper we show that accuracies reported on any dataset are correlated and invoke the appropriate statistical technique --- the paired model --- to compare the accuracies of two recognition systems. Theoretically we show that this method provides tighter confidence intervals than methods used in OCR and computer vision literature. We also propose a new visualization method, which we call the accuracy scatter plot, for providing a visual summary of performance results. This method summarizes the accuracy comparisons on the entire cor... IR carvalho00genetic A Genetic Algorithm-Based Solution for the Problem of Small Disjuncts . In essence, small disjuncts are rules covering a small number of examples. Hence, these rules are usually error-prone, which contributes to a decrease in predictive accuracy. The problem is particularly serious because, although each small disjuncts covers few examples, the set of small disjuncts can cover a large number of examples. This paper proposes a solution to the problem of discovering accurate small-disjunct rules based on genetic algorithms. The basic idea of our method is to use a hybrid decision tree / genetic algorithm approach for classification. More precisely, examples belonging to large disjuncts are classified by rules produced by a decision-tree algorithm, while examples belonging to small disjuncts are classified by a new genetic algorithm, particularly designed for discovering small-disjunct rules. 1 Introduction In the context of the well-known classification task of data mining, the discovered knowledge is often expressed as a set of IF-THEN predict... ML pouzol01from From Declarative Signatures to Misuse IDS In many existing misuse intrusion detection systems, intrusion signatures are very close to the detection algorithms. As a consequence, they contain too many cumbersome details. Recent work have proposed declarative signature languages that raise the level of abstraction when writing signatures. However, these languages do not always come with operational support. In this article, we show how to transform such declarative signatures into operational ones. This process points out several technical details which must be considered with care when performing the translation by hand, but which can be systematically handled. AI 300668 Building a Large Location Table to Find Replicas of Physics Objects The problem of building a large location table for physics objects occurs within a number of planned physics data management systems, like those that control reclustering and wide-area replication. To satisfy their e#ciency goals, these systems have to make local or remote replicas of individual physics objects, which contain raw or reconstructed data for a single event, rather than replicas of large run or ntuple files. This replication implies the use of a table to resolve the logical, location independent object descriptor into a physical location where an object replica can be found. For modern physics experiments the table needs to scale to at least some 10 10 objects. We argue that such a table can be e#ciently implemented by limiting the freedom of lookup operations, and by exploiting some specific properties of the physics data model. One specific viable implementation is discussed. Key words: Object location table, object-oriented databases, object clustering, object re-cl... DB cleary00generating Generating a Topically Focused VirtualReality Internet Surveys highlight that Internet users are frequently frustrated by failing to locate useful information, and by difficulty in browsing anarchically linked web-structures. We present a new Internet browsing application (called VR-net) that addresses these problems. It first identifies semantic domains consisting of tightly interconnected web-page groupings. The second part populates a 3D virtual world with these information sources, representing all relevant pages plus appropriate structural relations. Users can then easily browse through around a semantically focused virtual library. 1 Introduction The Internet is a probably the most significant global information resource ever created, allowing access to an almost unlimited amount of information. In this paper we describe two inter-related difficulties suffered by Internet users, and their combined influence on web use. We then introduce an integrated "search and browse" solution tool that directly tackles both issues. We also examin... IR wu00adaptive An Adaptive Self-Organizing Color Segmentation Algorithm with Application to Robust Real-time Human Hand Localization In Proc. Asian Conf. on Computer Vision, Taiwan, 2000 This paper describes an adaptive self-organizing color segmentation algorithm and a transductive learning algorithm used to localize human hand in video sequences. The color distribution at each time frame is approximated by the proposed 1-D self-organizing map (SOM), in which schemes of growing, pruning and merging are facilitated to find an appropriate number of color cluster automatically. Due to the dynamic backgrounds and changing lighting conditions, the distribution of color over time may not be stationary. An algorithm of SOM transduction is proposed to learn the nonstationary color distribution in HSI color space by combining supervised and unsupervised learning paradigms. Color cue and motion cue are integrated in the localization system, in which motion cue is employed to focus the attention of the system. This approach is also applied to other tasks such as human face tracking and color indexing. Our localization system... HCI bergmark02collection Collection Synthesis The invention of the hyperlink and the HTTP transmission protocol caused an amazing new structure to appear on the Internet -- the World Wide Web. With the Web, there came spiders, robots, and Web crawlers, which go from one link to the next checking Web health, ferreting out information and resources, and imposing organization on the huge collection of information (and dross) residing on the net. This paper reports on the use of one such crawler to synthesize document collections on various topics in science, mathematics, engineering and technology. Such collections could be part of a digital library. IR 32986 Min-Wise Independent Permutations We define and study the notion of min-wise independent families of permutations. We say that F # S n is min-wise independent if for any set X # [n] and any x # X, when # is chosen at random in F we have Pr(min{#(X)} = #(x)) = 1 |X| . In other words we require that all the elements of any fixed set X have an equal chance to become the minimum element of the image of X under #. Our research was motivated by the fact that such a family (under some relaxations) is essential to the algorithm used in practice by the AltaVista web index software to detect and filter near-duplicate documents. However, in the course of our investigation we have discovered interesting and challenging theoretical questions related to this concept -- we present the solutions to some of them and we list the rest as open problems. # Digital SRC, 130 Lytton Avenue, Palo Alto, CA 94301, USA. E-mail: broder@pa.dec.com. + Computer Science Department, Stanford University, CA 94305, USA. E-mail: moses@cs.stan... AI delgado99multiagent A Multiagent Architecture For Fuzzy Modeling In this paper a hybrid learning system that combines different fuzzy modeling techniques is being investigated. In order to implement the different methods, we propose the use of intelligent agents, which collaborate by means of a multiagent architecture. Agents 28223 Probabilistic Deduction with Conditional Constraints over Basic Events We study the problem of probabilistic deduction with conditional constraints over basic events. We show that globally complete probabilistic deduction with conditional constraints over basic events is NP-hard. We then concentrate on the special case of probabilistic deduction in conditional constraint trees. We elaborate very efficient techniques for globally complete probabilistic deduction. In detail, for conditional constraint trees with point probabilities, we present a local approach to globally complete probabilistic deduction, which runs in linear time in the size of the conditional constraint trees. For conditional constraint trees with interval probabilities, we show that globally complete probabilistic deduction can be done in a global approach by solving nonlinear programs. We show how these nonlinear programs can be transformed into equivalent linear programs, which are solvable in polynomial time in the size of the conditional constraint trees. 1. Introduction Dealing wit... AI 360831 Exploring Brick-Based Navigation and Composition in an Augmented Reality . BUILD-IT is a planning tool based on computer vision technology, supporting complex planning and composition tasks. A group of people, seated around a table, interact with objects in a virtual scene using real bricks. A plan view of the scene is projected onto the table, where object manipulation takes place. A perspective view is projected on the wall. The views are set by virtual cameras, having spatial attributes like shift, rotation and zoom. However, planar interaction with bricks provides only position and rotation information. Object height control is equally constrained by planar interaction. The aim of this paper is to suggest methods and tools bridging the gap between planar interaction and three-dimensional control. To control camera attributes, active objects, with intelligent behaviour are introduced. To control object height, several real and virtual tools are suggested. Some of the solutions are based on metaphors, like window, sliding-ruler and floor. 1 I... HCI 202938 On2broker: Semantic-Based Access to Information Sources at the WWW On2broker provides brokering services to improve access to heterogeneous, distributed and semistructured information sources as they are presented in the World Wide Web. It relies on the use of ontologies to make explicit the semantics of web pages. In the paper we will discuss the general architecture and main components of On2broker and provide some application scenarios. 1. Introduction In the paper we describe a tool environment called On2broker 1 that processes information sources and content descriptions in HTML, XML, and RDF and that provides intelligent information retrieval, query answering and maintenance support. Central for our approach is the use of ontologies to describe background knowledge and to make explicit the semantics of web documents. Ontologies have been developed in the area of knowledge-based systems for structuring and reusing large bodies of knowledge (cf. CYC [Lenat, 1995], (KA)2 [Benjamins et al., 1998]). Ontologies are consensual and formal specificat... DB damiani97semantic Semantic Approaches to Structuring and Querying Web Sites In order to pose effective queries to Web sites, some form of site data model must be implicitly or explicitly shared by users. Many approaches try to compensate for the lack of such a common model by considering the hypertextual structure of Web sites; unfortunately, this structure has usually little to do with data semantics. In this paper a different technique is proposed that allows for both navigational and logical/conceptual description of Web sites. The data model is based on WG-log, a query language based on the graph-oriented database model of GOOD [Gys94] and G-log [Par95], which allows the description of data manipulation primitives via (sets of) graph(s). The WG-log description of a Web site schema is lexically based on standard hypermedia design languages, thus allowing for easy schema generation by current hypermedia authoring environments. The use of WG-log for queries allows graphic query construction with respect to both the navigational and the logical parts of schema... IR popescul00automatic Automatic Labeling of Document Clusters Automatically labeling document clusters with words which indicate their topics is difficult to do well. The most commonly used method, labeling with the most frequent words in the clusters, ends up using many words that are virtually void of descriptive power even after traditional stop words are removed. Another method, labeling with the most predictive words, often includes rather obscure words. We present two methods of labeling document clusters motivated by the model that words are generated by a hierarchy of mixture components of varying generality. The first method assumes existence of a document hierarchy (manually constructed or resulting from a hierarchical clustering algorithm) and uses a 2 test of significance to detect different word usage across categories in the hierarchy. The second method selects words which both occur frequently in a cluster and effectively discriminate the given cluster from the other clusters. We compare these methods on abstracts of documents sel... IR vdovjak01rdf RDF Based Architecture for Semantic Integration of Heterogeneous Information Sources . The proposed integration architecture aims at exploiting data semantics in order to provide a coherent and meaningful (with respect to a given conceptual model) view of the integrated heterogeneous information sources. The architecture is split into five separate layers to assure modularization, providing description, requirements, and interfaces for each. It favors the lazy retrieval paradigm over the data warehousing approach. The novelty of the architecture lies in the combination of semantic and on-demand driven retrieval. This line of attack offers several advantages but brings also challenges, both of which we discuss with respect to RDF, the architecture's underlying model. 1 Introduction, Background, and Related Work With the vast expansion of the World Wide Web during the last few years the integration of heterogeneous information sources has become a hot topic. A solution to this integration problem allows for the design of applications that provide a uniform access to dat... DB kodratoff00comparing Comparing Machine Learning and Knowledge Discovery in DataBases: An Application to Knowledge Discovery in Texts INTRODUCTION KDD is better known by the oversimplified name of Data Mining (DM). Actually, most academics are rather interested by DM which develops methods for extracting knowledge from a given set of data. Industrialists and experts should be more interested in KDD which comprises the whole process of data selection, data cleaning, transfer to a DM technique, applying the DM technique, validating the results of the DM technique, and finally interpreting them for the user. In general, this process is a cycle that improves under the criticism of the expert. Machine Learning (ML) and KDD have in common a very strong link : they both acknowledge the importance of induction as a normal way of thinking, while other scientific fields are reluctant to accept it, to say the least. We shall first explore this common point. We believe that this reluctance relies on a misuse of apparent contradictions inside the theory of confirmation, that is we shall revisit Hempel paradox in order t IR granlund01patternsupported A Pattern-Supported Approach to the User Interface Design Process Patterns describe generic solutions to common problems in context. Originating from the world of architecture, patterns have been used mostly in object-oriented programming and data analysis. The goal of HCI patterns is to create an inventory of solutions to help designers (and usability engineers) to resolve UI development problems that are common, difficult and frequently encountered. In this paper, we present our pattern-supported approach to user interface design in the context of information visualization. Using a concrete example from the telecommunications domain, we will focus on a task/subtask pattern to illustrate how knowledge about a task and an appropriate interaction design solution can be captured and communicated. 1 HCI friedman99learning Learning Probabilistic Relational Models A large portion of real-world data is stored in commercial relational database systems. In contrast, most statistical learning methods work only with "flat " data representations. Thus, to apply these methods, we are forced to convert our data into a flat form, thereby losing much of the relational structure present in our database. This paper builds on the recent work on probabilistic relational models (PRMs), and describes how to learn them from databases. PRMs allow the properties of an object to depend probabilistically both on other properties of that object and on properties of related objects. Although PRMs are significantly more expressive than standard models, such as Bayesian networks, we show how to extend well-known statistical methods for learning Bayesian networks to learn these models. We describe both parameter estimation and structure learning — the automatic induction of the dependency structure in a model. Moreover, we show how the learning procedure can exploit standard database retrieval techniques for efficient learning from large datasets. We present experimental results on both real and synthetic relational databases. 1 DB kemme00new A New Approach to Developing and Implementing Eager Database Replication Protocols Database replication is traditionally seen as a way to increase the availability and performance of distributed databases. Although a large number of protocols providing data consistency and fault-tolerance have been proposed, few of these ideas have ever been used in commercial products due to their complexity and performance implications. Instead, current products allow inconsistencies and often resort to centralized approaches which eliminates some of the advantages of replication. As an alternative, we propose a suite of replication protocols that addresses the main problems related to database replication. On the one hand, our protocols maintain data consistency and the same transactional semantics found in centralized systems. On the other hand, they provide flexibility and reasonable performance. To do so, our protocols take advantage of the rich semantics of group communication primitives and the relaxed isolation guarantees provided by most databases. This allows us to eliminate the possibility of deadlocks, reduce the message overhead and increase performance. A detailed simulation study shows the feasibility of the approach and the flexibility with which different types of bottlenecks can be circumvented. DB gotoh00language Language Model Adaptation .15> attempt to exploit longer distance dependencies. -- infer some notion of `topic' from text. -- compute topic dependent probability. 8th ELSNET summer school 2 Language Model Adaptation 26 July 2000 ' & $ % Adaptive Language Modelling Stage 1: automatic derivation of topic information from text. ffl loose definition of document: a unit of spoken (or written) data of a certain length that contains some topic(s), or content(s). ffl topic of a document (= long distance or document-wide statistics. ffl information retrieval (IR): `bag-of-words' model based on a histogram of weighted unigram frequencies. Stage 2: combination of global and topic-dependent text statistics. ffl mixture. ffl maximum entropy modelling. (ref) Jelinek (1997). ' & $ % Mixtur IR callan99automatic Automatic Discovery of Language Models for Text Databases The proliferation of text databases within large organizations and on the Internet makes it difficult for a person to know which databases to search. Given language models that describe the contents of each database, a database selection algorithm such as GlOSS can provide assistance by automatically selecting appropriate databases for an information need. Current practice is that each database provides its language model upon request, but this cooperative approach has important limitations. This paper demonstrates that cooperation is not required. Instead, the database selection service can construct its own language models by sampling database contents via the normal process of running queries and retrieving documents. Although random sampling is not possible, it can be approximated with carefully selected queries. This sampling approach avoids the limitations that characterize the cooperative approach, and also enables additional capabilities. Experimental results demonstrate th... IR drummond97using Using a Case Base of Surfaces to Speed-Up Reinforcement Learning . This paper demonstrates the exploitation of certain vision processing techniques to index into a case base of surfaces. The surfaces are the result of reinforcement learning and represent the optimum choice of actions to achieve some goal from anywhere in the state space. This paper shows how strong features that occur in the interaction of the system with its environment can be detected early in the learning process. Such features allow the system to identify when an identical, or very similar, task has been solved previously and to retrieve the relevant surface. This results in an orders of magnitude increase in learning rate. 1 Introduction One important research issue for case based learning is its combination with other learning methods. As Aamodt and Plaza [1] point out, generally the machine learning community aims to produce "a coherent framework, where each learning method fulfills a specific and distinct role in the system." This paper discusses one such approach, combinin... ML cacheda01superimposing Superimposing Codes Representing Hierarchical Information in Web directories In this article we describe how superimposed coding can be used to represent hierarchical information, which is especially useful in categorized information retrieval systems (for example, Web directories). Superimposed coding have been widely used in signature files in a rigid manner, but our approach is more flexible and powerful. The categorization is based on a directed acyclic graph and each document is assigned to one or more nodes, using superimposed coding we represent the categorization information of each document in a signature. In this paper we explain the superimposed coding theory and how this coding technique can be applied to more flexible environments. Furthermore, we realize an exhaustive analysis of the important factors that have repercussions on the performance of the system. Finally we expose the conclusions obtained from this article. IR bhalotia02keyword Keyword Searching and Browsing in Databases using BANKS With the growth of the Web, there has been a rapid increase in the number of users who need to access online databases without having a detailed knowledge of the schema or of query languages; even relatively simple query languages designed for non-experts are too complicated for them. We describe BANKS, a system which enables keyword-based search on relational databases, together with data and schema browsing. BANKS enables users to extract information in a simple manner without any knowledge of the schema or any need for writing complex queries. A user can get information by typing a few keywords, following hyperlinks, and interacting with controls on the displayed results. BANKS models tuples as nodes in a graph, connected by links induced by foreign key and other relationships. Answers to a query are modeled as rooted trees connecting tuples that match individual keywords in the query. Answers are ranked using a notion of proximity coupled with a notion of prestige of nodes based on inlinks, similar to techniques developed for Web search. We present an efficient heuristic algorithm for finding and ranking query results. 1. IR timm01from From Corporate Memories to Supply Web Memory Modern production has discovered knowledge as an additional factor of production and a new trend of research, development and implementation of corporate memory systems is arising. The global economy leads to tighter corporation relations between enterprises. Therefore the knowledge of one product does not exist in a single company but within participating companies respective the supply chain. A modern product centered knowledge management has to face the difficult task of the integration of distributed knowledge sources. This contribution states our interest in research on the integration of corporate memory. In a first step we are focusing on single products leading to supply chain memories. Further research and development will lead to supply web memory. 1 Agents ghani01combining Combining Labeled and Unlabeled Data for Text Classification With a Large Number of Categories A major concern with supervised learning techniques for text classification is that they often require a large number of labeled examples to learn accurately. One way to reduce the amount of labeled data required is to develop algorithms that can learn effectively from a small number of labeled examples augmented with a large number of unlabeled examples. In this paper, we develop a framework to incorporate unlabeled data in the Error-Correcting Output Coding (ECOC) setup by decomposing multiclass problems into multiple binary problems and then use Co-Training to learn the individual binary classification problems. We show that our method is especially useful for classification tasks involving a large number of categories where Co-training doesn't perform very well by itself and when combined with ECOC, outperforms several other algorithms that combine labeled and unlabeled data for text classification. 1 IR 3489 Mixed Depth Representations for Dialog Processing We describe our work on developing a general purpose tutoring system that will allow students to practice their decision-making skills in a number of domains. The tutoring system, B2, supports mixed-initiative natural language interaction. The natural language processing and knowledge representation components are also general purpose|which leads to a tradeo between the limitations of super cial processing and syntactic representations and the di culty of deeper methods and conceptual representations. Students ' utterances may be short and ambiguous, requiring extensive reasoning about the domain or the discourse model to fully resolve. However, full disambiguation is rarely necessary. Our solution is to use a mixed-depth representation, one that encodes syntactic and conceptual information in the same structure. As a result, we can use the same representation framework to produce a detailed representation of requests (which tend to be well-speci ed) and to produce a partial representation of questions (which tend to require more inference about the context). Moreover, the representations use the same knowledge representation framework that is used to reason about discourse processing and domain information|so that the system can reason with (and about) the utterances, if necessary. AI michaud99representation Representation of behavioral history forl earningin nonstationary conditions A robot having to operate in nonstationary conditions needs to learn how to modify its control policy to adapt to the changing dynamics of the environment. Using the behavior-based approach to manage the interactions between the robot and its environment, we propose a method that models these interactions and adapts the selection of behaviors according to the history of behavior use. The learning and the use of this "Interaction Model " are validated using a vision- and sonar-based Pioneer I robot in the context of a multi-robot foraging task. Results show the effectiveness of the approach in taking advantage of any regularities experienced in the world, leading to fast and adaptable specialization for the learning robot. 1 AI muslea01hierarchical Hierarchical Wrapper Induction for Semistructured Information Sources With the tremendous amount of information that becomes available on the Web on a daily basis, the abilitytoquickly develop information agents has become a crucial problem. A vital componentofanyWeb-based information agent is a set of wrappers that can extract the relevant data from semistructured information sources. Our novel approach to wrapper induction is based on the idea of hierarchical information extraction, which turns the hard problem of extracting data from an arbitrarily complex documentinto a series of simpler extraction tasks. We introduce an inductive algorithm, stalker, that generates high accuracy extraction rules based on user-labeled training examples. Labeling the training data represents the major bottleneck in using wrapper induction techniques, and our experimental results showthatstalker requires up to two orders of magnitude fewer examples than other algorithms. Furthermore, can wrap information sources that could not be wrapped by existing inductivetechniques. IR weng98visionguided Vision-Guided Navigation Using SHOSLIF . This paper presents an unconventional approach to vision-guided autonomous navigation. The system recalls information about scenes and navigational experience using content-based retrieval from a visual database. To achieve a high applicability to various road types, we do not impose a priori scene features, such as road edges, that the system must use. But rather, the system automatically derives features from images during supervised learning. To accomplish this, the system uses principle component analysis and linear discriminant analysis to automatically derive the most expressive features (MEF) for scene reconstruction or the most discriminating features (MDF) for scene classi cation. These features best describe or classify the population of the scenes and approximate complex decision regions using piecewise linear boundaries up to a desired accuracy. A new self-organizing scheme called recursive partition tree (RPT) is used for automatic construction of a vision-and-control da... ML 428001 Learning and Tracking Human Motion Using Functional Analysis We present a method for the modeling and tracking of human motion using a sequence of 2D video images. Our analysis is divided in two parts: statistical learning and Bayesian tracking. First, we estimate a statistical model of typical activities from a large set of 3D human motion data. For this purpose, the human body is represented as a set of articulated cylinders and the evolution of a particular joint angle is described by a time-series. Specifically, we consider periodic motion such as “walking ” in this work, and we develop a new set of tools that allows for the automatic segmentation of the training data into a sequence of identical “motion cycles”. Then we compute the mean and the principal components of these cycles using a new algorithm to account for missing information and to enforce smooth transitions between different cycles. The learned temporal model provides a prior probability distribution over human motions which is used for tracking. We adopt a Bayesian perspective and approximate the posterior distribution of the body parameters using a particle filter. The resulting algorithm is able to track human subjects in monocular video sequences and to recover their 3D motion in complex unknown environments. 1 HCI 450055 Engineering Mobile-agent Applications via Context-dependent Coordination The design and development of Internet applications can take advantage of a paradigm based on autonomous and mobile agents. However, mobility introduces peculiar coordination problems in agent-based Internet applications. First, it suggests the exploitation of an infrastructure based on a multiplicity of local interaction spaces. Second, it may require coordination activities to be adapted both to the characteristics of the execution environment where they occur and to the needs of the application to which the coordinating agents belong. In this context, this paper introduces the concept of context-dependent coordination based on programmable interaction spaces. On the one hand, interaction spaces associated to different execution environments may be independently programmed so as to lead to differentiated, environment-dependent, behaviors. On the other hand, agents can program the interaction spaces of the visited execution environments to obtain an application-dependent behavior of the interaction spaces themselves. Several examples show how an infrastructure enforcing context-dependent coordination can be effectively exploited to simplify and make more modular the design of Internet applications based on mobile agents. In addition, the MARS coordination infrastructure is presented as an example of a system in which the concept of context-dependent coordination has found a clean and efficient implementation. Agents 38408 Continuous Categories For a Mobile Robot Autonomous agents make frequent use of knowledge in the form of categories --- categories of objects, human gestures, web pages, and so on. This paper describes a way for agents to learn such categories for themselves through interaction with the environment. In particular, the learning algorithm transforms raw sensor readings into clusters of time series that have predictive value to the agent. We address several issues related to the use of an uninterpreted sensory apparatus and show specific examples where a Pioneer 1 mobile robot interacts with objects in a cluttered laboratory setting. Introduction "There is nothing more basic than categorization to our thought, perception, action, and speech" (Lakoff 1987). For autonomous agents, categories often appear as abstractions of raw sensor readings that provide a means for recognizing circumstances and predicting effects of actions. For example, such categories play an important role for a mobile robot that navigates around obstacles ... AI wolter00spatiotemporal Spatio-temporal representation and reasoning based on RCC-8 this paper is to introduce a hierarchy of languages intended for qualitative spatio-temporal representation and reasoning, provide these languages with topological temporal semantics, construct effective reasoning algorithms, and estimate their computational complexity. DB joachims01statistical A Statistical Learning Model of Text Classification for Support Vector Machines This paper develops a theoretical learning model of text classification for Support Vector Machines (SVMs). It connects the statistical properties of text-classification tasks with the generalization performance of a SVM in a quantitative way. Unlike conventional approaches to learning text classifiers, which rely primarily on empirical evidence, this model explains why and when SVMs perform well for text classification. In particular, it addresses the following questions: Why can support vector machines handle the large feature spaces in text classification effectively? How is this related to the statistical properties of text? What are sufficient conditions for applying SVMs to text-classification problems successfully? IR vaneijk01generalised Generalised Object-Oriented Concepts for Inter-Agent Communication . In this paper, we describe a framework to program open societies of concurrently operating agents. The agents maintain a subjective theory about their environment and interact with each other via a communication mechanism suited for the exchange of information, which is a generalisation of the traditional rendez-vous communication mechanism from the object-oriented programming paradigm. Moreover, following object-oriented programming, agents are grouped into agent classes according to their particular characteristics; viz. the program that governs their behaviour, the language they employ to represent information and most interestingly the questions they can be asked to answer. We give and operational model of the programming language in terms of a transition system for the formal derivation of computations of multi-agent programs. 1 Introduction The field of multi-agent systems is a rapidly growing research area. Although in this field there is no real consensus on what ... Agents bergamaschi98semantic A Semantic Approach to Information Integration: the MOMIS project this paper, we propose a semantic approach to the integration of heterogeneous information. The approach follows the semantic paradigm, in that conceptual schemata of an involved source are considered, and a common data model (ODM I 3) and language (ODL I 3) are adopted to describe sharable information. ODM I 3 and ODL I 3 are defined as a subset of the corresponding ODMG-93 [13] ODM and ODL. A Description Logics ocdl (object description language with constraints [6]) is used as a DB 460350 The Intelligent Surfer: Probabilistic Combination of Link and Content Information in PageRank The PageRank algorithm, used in the Google search engine, greatly improves the results of Web search by taking into account the link structure of the Web. PageRank assigns to a page a score proportional to the number of times a random surfer would visit that page, if it surfed indefinitely from page to page, following all outlinks from a page with equal probability. We propose to improve PageRank by using a more intelligent surfer, one that is guided by a probabilistic model of the relevance of a page to a query. Efficient execution of our algorithm at query time is made possible by precomputing at crawl time (and thus once for all queries) the necessary terms. Experiments on two large subsets of the Web indicate that our algorithm significantly outperforms PageRank in the (human -rated) quality of the pages returned, while remaining efficient enough to be used in today's large search engines. IR samtani98recent Recent Advances and Research Problems in Data Warehousing . In the recent years, the database community has witnessed the emergence of a new technology, namely data warehousing. A data warehouse is a global repository that stores pre-processed queries on data which resides in multiple, possibly heterogeneous, operational or legacy sources. The information stored in the data warehouse can be easily and efficiently accessed for making effective decisions. The On-Line Analytical Processing (OLAP) tools access data from the data warehouse for complex data analysis, such as multidimensional data analysis, and decision support activities. Current research has lead to new developments in all aspects of data warehousing, however, there are still a number of problems that need to be solved for making data warehousing effective. In this paper, we discuss recent developments in data warehouse modelling, view maintenance, and parallel query processing. A number of technical issues for exploratory research are presented and possible solutions are discusse... DB 198731 Learning Environmental Features for Pose Estimation We present a method for learning a set of environmental features which are useful for pose estimation. The landmark learning mechanism is designed to be applicable to a wide range of environments, and generalized for di#erent sensing modilities. In the context of computer vision, each landmark is detected as a local extremum of a measure of distinctiveness and represented by an appearance-based encoding which is exploited for matching. The set of obtained landmarks can be parameterized and then evaluated in terms of their utility for the task at hand. The method is used to motivate a general approach to task-oriented sensor fusion. We present experimental evidence that demonstrates the utility of the method. 1 Introduction In this paper, we develop an approach to sensorbased robot localization by learning a set of recognizable features in the robot's environment. In particular, we consider the problem of learning a set of image-domain landmarks from a set of di#erent views of a scene. ... Agents 508492 SCANMail: a voicemail interface that makes speech browsable, readable and searchable Increasing amounts of public, corporate, and private speech data are now available on-line. These are limited in their usefulness, however, by the lack of tools to permit their browsing and search. The goal of our research is to provide tools to overcome the inherent difficulties of speech access, by supporting visual scanning, search, and information extraction. We describe a novel principle for the design of UIs to speech data: What You See Is Almost What You Hear (WYSIAWYH). In WYSIAWYH, automatic speech recognition (ASR) generates a transcript of the speech data. The transcript is then used as a visual analogue to that underlying data. A graphical user interface allows users to visually scan, read, annotate and search these transcripts. Users can also use the transcript to access and play specific regions of the underlying message. We first summarize previous studies of voicemail usage that motivated the WYSIAWYH principle, and describe a voicemail UI, SCANMail, that embodies WYSIAWYH. We report on a laboratory experiment and an 18 user, two month field trial evaluation. SCANMail outperformed a state of the art voicemail system on core voicemail tasks. This was attributable to SCANMail's support for visual scanning, search and information extraction. While the ASR transcripts contain errors, they nevertheless improve the efficiency of voicemail processing. Transcripts either provide enough information for users to extract key points or to navigate to important regions of the underlying speech, which they can then play directly. IR ambroszkiewicz98team Team Formation by Self-Interested Mobile Agents . A process of team formation by autonomous agents in a distributed environment is presented. Since the environment is distributed, there are serious problems with communication and consistent decision making inside a team. To deal with these problems, the standard technique of token passing in a computer network is applied. The passing cycle of the token serves as the communication route. It assures consistent decision making inside the team maintaining its organizational integrity. On the other hand it constitutes a component of the plan of the cooperative work performed by a complete team. Two algorithms for team formation are given. The first one is based on simple self-interested agents that still can be viewed as reactive agents (see [14]) although augmented with knowledge, goal, and cooperation mechanisms. The second one is based on sophisticated self-interested agents. Moreover, the algorithm based on fully cooperative agents, which is an adaptation of the static ... Agents bolotov99clausal A Clausal Resolution Method for CTL Branching-Time Temporal Logic In this paper we extend our clausal resolution method for linear time temporal logics to a branching-time framework. Thus, we propose an efficient deductive method useful in a variety of applications requiring an expressive branching-time temporal logic in AI. The branching-time temporal logic considered is Computation Tree Logic (CTL), often regarded as the simplest useful logic of this class. The key elements of the resolution method, namely the normal form, the concept of step resolution and a novel temporal resolution rule, are introduced and justified with respect to this logic. A completeness argument is provided, together with some examples of the use of the temporal resolution method. Finally, we consider future work, in particular the extension of the method yet further, to Extended CTL (ECTL), which is CTL extended with fairness operators, and CTL , the most powerful logic of this class. We will also outline possible implementation of the approach by adapting techniques de... DB middleton02exploiting Exploiting Synergy Between Ontologies and Recommender Systems Recommender systems learn about user preferences over time, automatically finding things of similar interest. This reduces the burden of creating explicit queries. Recommender systems do, however, suffer from cold-start problems where no initial information is available early on upon which to base recommendations. Semantic knowledge structures, such as ontologies, can provide valuable domain knowledge and user information. However, acquiring such knowledge and keeping it up to date is not a trivial task and user interests are particularly difficult to acquire and maintain. This paper investigates the synergy between a web-based research paper recommender system and an ontology containing information automatically extracted from departmental databases available on the web. The ontology is used to address the recommender systems cold-start problem. The recommender system addresses the ontology’s interest-acquisition problem. An empirical evaluation of this approach is conducted and the performance of the integrated systems measured. IR smyth01data Data Mining At The Interface Of Computer Science And Statistics This chapter is written for computer scientists, engineers, mathematicians, and scientists who wish to gain a better understanding of the role of statistical thinking in modern data mining. Data mining has attracted considerable attention both in the research and commercial arenas in recent years, involving the application of a variety of techniques from both computer science and statistics. The chapter discusses how computer scientists and statisticians approach data from different but complementary viewpoints and highlights the fundamental differences between statistical and computational views of data mining. In doing so we review the historical importance of statistical contributions to machine learning and data mining, including neural networks, graphical models, and flexible predictive modeling. The primary conclusion is that closer integration of computational methods with statistical thinking is likely to become increasingly important in data mining applications. Keywords: Data mining, statistics, pattern recognition, transaction data, correlation. 1. ML 510100 On the Robustness of some Cryptographic Protocols for Mobile Agent Protection Mobile agent security is still a young discipline and most naturally, the focus up to the time of writing was on inventing new cryptographic protocols for securing various aspects of mobile agents. However, past experience shows that protocols can be flawed, and flaws in protocols can remain unnoticed for a long period of time. The game of breaking and fixing protocols is a necessary evolutionary process that leads to a better understanding of the underlying problems and ultimately to more robust and secure systems. Although, to the best of our knowledge, little work has been published on breaking protocols for mobile agents, it is inconceivable that the multitude of protocols proposed so far are all flawless. As it turns out, the opposite is true. We identify flaws in protocols proposed by Corradi et al., Karjoth et al., and Karnik et al., including protocols based on secure co-processors. Agents yu01indexing Indexing the Distance: An Efficient Method to KNN Processing In this paper, we present an efficient method, called iDistance, for K-nearest neighbor (KNN) search in a high-dimensional space. iDistance partitions the data and selects a reference point for each partition. The data in each cluster are transformed into a single dimensional space based on their similarity with respect to a reference point. This allows the points to be indexed using a B + -tree structure and KNN search be performed using one-dimensional range search. The choice of partition and reference point provides the iDistance technique with degrees of freedom most other techniques do not have. We describe how appropriate choices here can effectively adapt the index structure to the data distribution. We conducted extensive experiments to evaluate the iDistance technique, and report results demonstrating its effectiveness. DB sim00to To Boldly Go: Bayesian Exploration for Mobile Robots This work addresses the problem of robot exploration. That is, the task of automatically learning a map of the environment which is useful for mobile robot navigation and localization. The exploration mechanism is intended to be applicable to an arbitrary environment, and is independent of the particular representation of the world. We take an information-theoretic approach and avoid the use of arbitrary heuristics. Preliminary results are presented and we discuss future directions for investigation. ML 132458 Equal Time for Data on the Internet with WebSemantics . Many collections of scientific data in particular disciplines are available today around the world. Much of this data conforms to some agreed upon standard for data exchange, i.e., a standard schema and its semantics. However, sharing this data among a global community of users is still difficult because of a lack of standards for the following necessary functions: (i) data providers need a standard for describing or publishing available sources of data; (ii) data administrators need a standard for discovering the published data and (iii) users need a standard for accessing this discovered data. This paper describes a prototype implementation of a system, WebSemantics, that accomplishes the above tasks. We describe an architecture and protocols for the publication, discovery and access to scientific data. We define a language for discovering sources and querying the data in these sources, and we provide a formal semantics for this language. 1 Introduction Recently, many standardized... IR 381995 A Practical Approach for Recovery of Evicted Variables SRC’s charter is to advance the state of the art in computer systems by doing basic and applied research in support of our company’s business objectives. Our interests and projects span scalable systems (including hardware, networking, distributed systems, and programming-language technology), the Internet (including the Web, e-commerce, and information retrieval), and human/computer interaction (including user-interface technology, computer-based appliances, and mobile computing). SRC was established in 1984 by Digital Equipment Corporation. We test the value of our ideas by building hardware and software prototypes and assessing their utility in realistic settings. Interesting systems are too complex to be evaluated solely in the abstract; practical use enables us to investigate their properties in depth. This experience is useful in the short term in refining our designs and invaluable in the long term in advancing our knowledge. Most of the major advances in information systems have come through this approach, including personal computing, distributed systems, and the Internet. We also perform complementary work of a more mathematical character. Some of HCI 19461 Intelligent Diagnosis Systems This paper examines and compares several different approaches to the design of Intelligent Systems for Diagnosis and Advising applications. These include expert systems or knowledge-based systems, case-based reasoning systems, truth (or reason) maintenance systems, statistical pattern classification systems, decision trees, and artificial neural networks (or connectionist systems). The key aspects of each approach are demonstrated through the design of a system for a simple automobile fault diagnosis task. The paper also discusses the domain characteristics and design and performance requirements that influence the choice of a specific technique (or a combination of techniques) for a given application. 1 Introduction The last few decades have seen a proliferation of intelligent systems for diagnosis, advising, forecasting, and related applications (Dean et al., 1995; Durkin, 1994; Ginsberg, 1993; Luger & Stubblefield, 1993; Puppe, 1993; Rich & Knight, 1991; Russell & Norvig, 1995; Ste... ML smith00conversation Conversation Trees and Threaded Chats Chat programs and instant messaging services are increasingly popular among Internet users. However, basic issues with the interfaces and data structures of most forms of chat limit their utility for use in formal interactions (like group meetings) and decision-making tasks. In this paper, we discuss Threaded Text Chat, a program designed to address some of the deficiencies of current chat programs. Standard forms of chat introduce ambiguity into interaction in a number of ways, most profoundly by rupturing connections between turns and replies. Threaded Chat presents a solution to this problem by actively supporting the basic turn-taking structure of human conversation. While the solution introduces interface design challenges of its own, usability studies show that users' patterns of interaction in Threaded Chat are equally effective, but different (and possibly more efficient) than standard chat programs. Keywords Chat programs, turn-taking, conversation, computer mediated communi... HCI srinivasan02target Target Seeking Crawlers and their Topical Performance Topic driven crawlers can complement search engines by targeting relevant portions of the Web. A topic driven crawler must exploit the information available about the topic and its underlying context. In this paper we extend our previous research on the design and evaluation of topic driven crawlers by comparing seven different crawlers on a harder problem, namely, seeking highly relevant target pages. We find that exploration is an important aspect of a crawling strategy. We also study how the performance of crawler strategies depends on a number of topical characteristics based on notions of topic generality, cohesiveness, and authoritativeness. Our results reveal that topic generality is an obstacle for most crawlers, that three crawlers tend to perform better when the target pages are clustered together, and that two of these also display better performance when topic targets are highly authoritative. IR hohl99nexus Nexus - An Open Global Infrastructure for Spatial-Aware Applications Due to the lack of a generic platform for location- and spatial-aware systems, many basic services have to be reimplemented in each application that uses spatial-awareness. A cooperation among different applications is also difficult to achieve without a common platform. In this paper we present a platform that solves these problems. It provides an infrastructure that is based on computer models of regions of the physical world, which are augmented by virtual objects. We show how virtual objects make the integration of existing information systems and services in spatial-aware systems easier. Furthermore, our platform supports interactions between the computer models and the real world and integrates single models in a global "Augmented World". Contents 1 Introduction 3 2 General Idea 4 2.1 Augmented Areas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2.2 Augmented World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 3 Example Scenario 6 4 Require... HCI brazier99compositional Compositional Design and Reuse of a Generic Agent Model This paper introduces a formally specified design of a compositional generic agent model (GAM). This agent model abstracts from specific application domains; it provides a unified formal definition of a model for weak agenthood. It can be (re)used as a template or pattern for a large variety of agent types and application domain types. The model was designed on the basis of experiences in a number of application domains. The compositional development method DESIRE was used to design the agent model GAM at a conceptual and logical level. It serves as a unified, precisely defined coneptual structure which can be refined by specialisation and instantiation to a large variety of other, more specific agents. To illustrate reuse of this agent model, specialisation and instantiation to model co-operative information gathering agents is described in depth. Moreover, it is shown how GAM can be used to describe in a unified and hence more comparable manner a large number of agent architectures from the literature. Agents 532267 A Comparison of Decision Making Criteria and Optimization Methods for Active Robotic Sensing This work presents a comparison of decision making criteria and optimization methods for active sensing in robotics. Active sensing incorporates the following aspects: (i) where to position sensors, and (ii) how to make decisions for next actions, in order to maximize information gain and minimize costs. We concentrate on the second aspect: "Where should the robot move at the next time step?". Pros and cons of the most often used statistical decision making strategies are discussed. ML surmann00five A Five Layer Sensor Architecture for Autonomous Robots in Indoor Environments Autonomous mobile service robots for transportation tasks in indoor environments e.g. multistory buildings, have to act in normal dynamic environments but with a huge number of components. The robots total repertoire of skills is high according to the complexity of the building and its respective task. Difficult tasks can only be achieved on the base by immediate sensing of the environment. This paper describes a five layer sensor architecture with an integrated world model for multistory buildings. In contrast to grid based approaches we use a feature based approach. The sensor architecture as well as the evaluation modules of the sensor data are based on natural landmarks. The key features of the sensor architecture are reuseability, modularity and portability to other multistory buildings as well as extendibility with different sensors. Keywords---robot architectures, sensor architectures, collision avoidance and sensor-based control, robot sensing and data fusion, behavior-based robotics I. ML andre99integrating Integrating Models of Personality and Emotions into Lifelike Characters . A growing number of research projects in academia and industry have recently started to develop lifelike agents as a new metaphor for highly personalised human-machine communication. A strong argument in favour of using such characters in the interface is the fact that they make humancomputer interaction more enjoyable and allow for communication styles common in human-human dialogue. In this paper we discuss three ongoing projects that use personality and emotions to address different aspects of the affective agent-user interface: (a) Puppet uses affect to teach children how the different emotional states can change or modify a character's behaviour, and how physical and verbal actions in social interactions can induce emotions in others; (b) the Inhabited Market Place uses affect to tailor the roles of actors in a virtual market place; and (c) Presence uses affect to enhance the believability of a virtual character, and produce a more natural conversational manner. Int... HCI coellocoello99updated An Updated Survey of Evolutionary Multiobjective Optimization Techniques: State of the Art and Future Trends This paper reviews some of the most popular evolutionary multiobjective optimization techniques currently reported in the literature, indicating some of their main applications, their advantages, disadvantages, and degree of aplicability. Finally, some of the most promising areas of future research are briefly discussed. 1 Introduction Since the pioneering work of Rosenberg in the late 1960s regarding the possibility of using genetic-based search to deal with multiple objectives [1], this new area of research (now called Evolutionary Multi-Objective Optimization, or EMOO for short) has grown considerably as indicates the notable increment (mainly in the last 15 years) of technical papers in international conferences and peer-reviewed journals, special sessions in international conferences and interest groups in the Internet 1 . Multiobjective optimization is with no doubt a very important research topic both for scientists and engineers, not only because of the multiobjective nature... ML czajkowski98jres JRes: A Resource Accounting Interface for Java With the spread of the Internet the computing model on server systems is undergoing several important changes. Recent research ideas concerning dynamic operating system extensibility are finding their way into the commercial domain, resulting in designs of extensible databases and Web servers. In addition, both ordinary users and service providers must deal with untrusted downloadable executable code of unknown origin and intentions. Across the board, Java has emerged as the language of choice for Internet-oriented software. We argue that, in order to realize its full potential in applications dealing with untrusted code, Java needs a flexible resource accounting interface. The design and prototype implementation of such an interface --- JRes --- is presented in this paper. The interface allows to account for heap memory, CPU time, and network resources consumed by individual threads or groups of threads. JRes allows limits to be set on resources available to threads and it can invoke... IR lacorte99agent An Agent Based Framework for Mobile Users User mobility together with an easy access to distributed resources is one of the greatest challenge to be faced in the future years. At the same time, agent technology is seen as a very promising approach to deal with distributed computing and user mobility. In this paper an agent-based strategy for support of mobile users is presented. It is based on a mobile agent platform developed at the University of Catania, which has been enhanced in order to allow the user to access network services in a mobile environment. Main functionalities and architecure of the above platform are described. 1 Introduction The quick expansion of wireless communication technologies and of portable computing devices, has made mobile computing more and more important. The user wishes to access the information he/she needs at any moment, independently of the place where he/she is. The ever increasing computing power available in notebooks, makes them a valid working tool for the user who needs to move from ... Agents swindells00system System Lag Tests for Augmented and Virtual Environments We describe a simple technique for accurately calibrating the temporal lag in augmented and virtual environments within the Enhanced Virtual Hand Lab (EVHL), a collection of hardware and software to support research on goal-directed human hand motion. Lag is the sum of various delays in the data pipeline associated with sensing, processing, and displaying information from the physical world to produce an augmented or virtual world. Our main calibration technique uses a modified phonograph turntable to provide easily tracked periodic motion, reminiscent of the pendulum-based calibration technique of Liang, Shaw and Green. Measurements show a three-frame (50 ms) lag for the EVHL. A second technique, which uses a specialized analog sensor that is part of the EVHL, provides a "closed loop" calibration capable of sub-frame accuracy. Knowing the lag to sub-frame accuracy enables a predictive tracking scheme to compensate for the end-toend lag in the data pipeline. We describe both techniques and the EVHL environment in which they are used. HCI camacho01mapweb MAPWEB: Cooperation between Planning Agents and Web Agents This paper presents MAPWeb (MultiAgent Planning in the Web), a multiagent system for cooperative work among dierent intelligent software agents whose main goal is to solve user planning problems using the information stored in the World Wide Web (Web). MAPWeb is made of a heterogeneous mixture of intelligent agents whose main characteristics are cooperation, reasoning, and knowledge sharing. The architecture of MAPWeb uses four types of agents:UserAgents that are the bridge between the users and the system; ControlAgents (Manager and Coach Agents) that are responsible to manage the rest of agents; PlannerAgents that are able to solve planning problems; and nally WebAgents whose aim is to retrieve, represent and share information obtained from the Web. MAPWeb solves planning problems by means of cooperation between PlannerAgents and WebAgents. Instead of trying the PlannerAgent to solve the whole planning problem, the PlannerAgent focuses on a less restricted (and therefore easier to solve) problem (what we call an abstract problem) and cooperates with the WebAgents to validate and complete abstract solutions. In order for cooperation to take place, a common language and data structures have also been dened. Categories and Subject Descriptors H.3.5 [Online Information Services]: Data sharing, Webbased services; I.2 [Articial Intelligence]; I.2.6 [Learning]: Knowledge acquisition; I.2.8 [Problem Solving]: Planning; I.2.11 [Distributed Articial Intelligence]: Intelligent agents, Multi-Agent Systems, Web agents Keywords Information System, Agent Architecture, Multi-Agent Systems, Web Agents, Intelligent Agents, Planning. 1. Agents 269982 Audio-Visual Speaker Detection using Dynamic Bayesian Networks The development of human-computer interfaces poses a challenging problem: actions and intentions of different users have to be inferred from sequences of noisy and ambiguous sensory data. Temporal fusion of multiple sensors can be efficiently formulated using dynamic Bayesian networks (DBNs). DBN framework allows the power of statistical inference and learning to be combined with contextual knowledge of the problem. We demonstrate the use of DBNs in tackling the problem of audio/visual speaker detection. "Off-the-shelf" visual and audio sensors (face, skin, texture, mouth motion, and silence detectors) are optimally fused along with contextual information in a DBN architecture that infers instances when an individual is speaking. Results obtained in the setup of an actual human-machine interaction system (Genie Casino Kiosk) demonstrate superiority of our approach over that of static, context-free fusion architecture. 1. Introduction Advanced human--computer interfaces increasingly r... AI cohen01learning Learning to Match and Cluster Entity Names Introduction Information retrieval is, in large part, the study of methods for assessing the similarity of pairs of documents. Document similarity metrics have been used for many tasks including ad hoc document retrieval, text classification [YC1994], and summarization [GC1998,SSMB1997]. Another problem area in which similarity metrics are central is record linkage (e.g., [KA1985]), where one wishes to determine if two database records taken from different source databases refer to the same entity. For instance, one might wish to determine if two database records from two different hospitals, each containing a patient's name, address, and insurance information, refer to the same person; as another example, one might wish to determine if two bibliography records, each containing a paper title, list of authors, and journal name, refer to the same publication. In both of these examples (and in many other practical cases) most of the record fields IR parsons99approach An approach to using degrees of belief in BDI agents : The past few years have seen a rise in the popularity of the use of mentalistic attitudes such as beliefs, desires and intentions to describe intelligent agents. Many of the models which formalise such attitudes do not admit degrees of belief, desire and intention. We see this as an understandable simplification, but as a simplification which means that the resulting systems cannot take account of much of the useful information which helps to guide human reasoning about the world. This paper starts to develop a more sophisticated system based upon an existing formal model of these mental attributes. 1 Introduction In the past few years there has been a lot of attention given to building formal models of autonomous software agents; pieces of software which operate to some extent independently of human intervention and which therefore may be considered to have their own goals and the ability to determine how to achieve those goals. Many of these formal models are based on the use of ... Agents huget02model Model Checking Agent UML Protocol Diagrams Agents in multiagent systems use protocols in order to exchange messages and to coordinate together. Since agents and objects are not exactly the same, designers do not use directly communication protocols used in distributed systems but a new type called interaction protocols encompassing agent features such as richer messages and the ability to cooperate and to coordinate. Obviously, designers consider formal description techniques used for communication protocols. New graphical modeling languages based on UML appeared several years ago. Agent UML is certainly the best known. Until now, no validation is given for Agent UML. The aim of this paper is to present how to model check Agent UML protocol diagrams. Agents 389181 A Fast Multi-Dimensional Algorithm for Drawing Large Graphs We present a novel hierarchical force-directed method for drawing large graphs. The algorithm produces a graph embedding in an Euclidean space E of any dimension. A two or three dimensional drawing of the graph is then obtained by projecting a higher-dimensional embedding into a two or three dimensional subspace of E. Projecting high-dimensional drawings onto two or three dimensions often results in drawings that are "smoother" and more symmetric. Among the other notable features of our approach are the utilization of a maximal independent set filtration of the set of vertices of a graph, a fast energy function minimization strategy, e#cient memory management, and an intelligent initial placement of vertices. Our implementation of the algorithm can draw graphs with tens of thousands of vertices using a negligible amount of memory in less than one minute on a mid-range PC. 1 Introduction Graphs are common in many applications, from data structures to networks, from software engineering... AI 494121 Some Considerations about Embodied Agents As computers are being more and more part of our world we feel the urgent need of proper user interface to interact with. The use of command lines typed on a keyboard are more and more obsolete, specially as computers are receiving so much attention from a large audience. The metaphor of face-to-face communication applied to human-computer interaction is finding a lot of attraction. Humans are used since they are born to communicate with others. Seeing faces, interpreting their expressions, understanding speech are all part of our development and growth. But face-to-face conversation is very complex as it involved a huge number of factors. We speak with our voice, but also with our hand, eye, face and body. Our gesture modifies, emphasizes, contradicts what we say by words. The production of speech and nonverbal behaviors work in parallel and not in antithesis. They seem to be two different forms (voice and body gestures) of the same process (speech). They add info Agents 11099 ifile: An Application of Machine Learning to E-Mail Filtering The rise of the World Wide Web and the ever-increasing amounts of machine-readable text has caused text classification to become a important aspect of machine learning. One specific application that has the potential to affect almost every user of the Internet is e-mail filtering. The WorldTalk Corporation estimates that over 60 million business people use e-mail [6]. Many more use e-mail purely on a personal basis and the pool of e-mail users is growing daily. And yet, automated techniques for learning to filter e-mail have yet to significantly affect the e-mail market. Here, I attack problems that plague practical e-mail ltering and suggest solutions that will bring us closer to the acceptance of using automated classification techniques to filter personal e-mail. I also present a filtering system, ifile, that is both effective and efficient, and which has been adapted to a popular e-mail client. Results are presented from a number of experiments and show that a system such as ifile could become a... DB marmasse99commotion comMotion: a context-aware communication system How many times have you gone to the grocery store but left the grocery list on the refrigerator door? Wouldn't it be more efficient to have a reminder to buy groceries and the shopping list delivered to you when you were in the vicinity of the store? We live in a world in which the information overload is part of our daily life. Many of us receive large quantities of email or voice mail messages. Yet many of these messages are relevant only in a particular context. We can use a system of reminders to keep up with all we have to do, but these reminders are often relevant only to a specific location. If reminders, to-do lists, messages and other information were delivered in the most timely and relevant context, part of the overload would be reduced. This paper describes comMotion, a context-aware communication system for a mobile or wearable computing platform. Keywords Mobile, ubiquitous and wearable computing; locationaware, context-aware applications. HCI 346149 Rewriting Logic as a Metalogical Framework A metalogical framework is a logic with an associated methodology that is used to represent other logics and to reason about their metalogical properties. We propose that logical frameworks can be good metalogical frameworks when their logics support reflective reasoning and their theories always have initial models. We present a concrete realization of this idea in rewriting logic. Theories in rewriting logic always have initial models and this logic supports reflective reasoning. This implies that inductive reasoning is valid when proving properties about the initial models of theories in rewriting logic, and that we can use reflection to reason at the metalevel about these properties. In fact, we can uniformly reflect induction principles for proving metatheorems about rewriting logic theories and their parameterized extensions. We show that this reflective methodology provides an effective framework for di erent, non-trivial, kinds of formal metatheoretic reasoning; one can, for examp... ML talim01controlling Controlling the Robots of Web Search Engines Robots are deployed by a Web search engine for collecting information from different Web servers in order to maintain the currency of its data base of Web pages. In this paper, we investigate the number of robots to be used by a search engine so as to maximize the currency of the data base without putting an unnecessary load on the network. We use a queueing model to represent the system. The arrivals to the queueing system are Web pages brought by the robots; service corresponds to the indexing of these pages. The objective is to find the number of robots, and thus the arrival rate of the queueing system, such that the indexing queue is neither starved nor saturated. For this, we consider a finite-buffer queueing system and define the cost function to be minimized as a weighted sum of the loss rate and the fraction of time that the system is empty. Both static and dynamic policies are considered. In the static setting the number of robots is held fixed; in the dynamic setting robots may be re... IR bertelsen99augmenting Augmenting reality in mobile substrates - On the design of computer support for process control The paper investigates augmented reality as a perspective on the design of computer support for process control in a distributed environment. Based on empirical studies of work in a wastewater treatment plant, three technical approaches on augmented reality --- augmenting the user; the object of work; and the environment --- are examined in terms of a collection of design scenarios .We conclude that these approaches when used, as metaphors rather than a consistent theoretical framework, may inform design of mobile support for process control work. Keywords: Augmented reality, process control, mobile computing, human-computer interaction, participatory design, workplace studies. BRT Keywords: AB, FA, FC, GA, HD Introduction Advanced technical (process) settings, such as modern wastewater treatment plants, are characterised by being highly distributed and dynamic. A possible strategy for supporting work in such settings is through the introduction of mobile technology. In this paper, ... HCI stanoi00reverse Reverse Nearest Neighbor Queries for Dynamic Databases In this paper we propose an algorithm for answering reverse nearest neighbor (RNN) queries, a problem formulated only recently. This class of queries is strongly related to that of nearest neighbor (NN) queries, although the two are not necessarily complementary. Unlike nearest neighbor queries, RNN queries find the set of database points that have the query point as the nearest neighbor. There is no other proposal we are aware of, that provides an algorithmic approach to answer RNN queries. The earlier approach for RNN queries ([KM99]) is based on the pre-computation of neighborhood information that is organized in terms of auxiliary data structures. It can be argued that the precomputation of the RNN information for all points in the database can be too restrictive. In the case of dynamic databases, insert and update operations are expensive and can lead to modifications of large parts of the auxiliary data structures. Also, answers to RNN queries for a set of data points depend on t... DB norrie02webbased Web-Based Integration of Printed and Digital Information The affordances of paper have ensured its retention as a key information medium, in spite of dramatic increases in the use of digital technologies for information storage, processing and delivery. Recent developments in paper, printing and wand technologies may lead to the widespread use of digitally augmented paper in the near future, thereby enabling the paper and digital worlds to be linked together. We are interested in using these technologies to achieve a true integration of printed and digital information sources such that users may browse freely back and forth between paper and digital resources. We present web-based server technologies that support this integration by allowing users to dynamically link areas of printed documents to objects of an application database. The server component is implemented using the eXtensible Information Management Architecture (XIMA) and is independent of the particular paper, printing and reader technologies used to realise the digitally augmented paper. The framework presented manages semantic information about application objects, documents, users, links and client devices and it supports universal client access. IR johns01how How Emotions and Personality Effect the Utility of Alternative Decisions: A Terrorist Target Selection Case Study The role of emotion modeling in the development of computerized agents has long been unclear. This is partially due to instability in the philosophical issues of the problem as psychologists struggle to build models for their own purposes, and partially due to the often-wide gap between these theories and that which can be implemented by an agent author. This paper describes an effort to use emotion models in part as a deep model of utility for use in decision theoretic agents. This allows for the creation of simulated forces capable of balancing a great deal of competing goals, and in doing so they behave, for better or for worse, in a more realistic manner. Agents kang01qrtdb qRTDB: Qos-Sensitive Real-Time Database Introduction Recently the demand for real-time database services is exploding. The applications requiring such services include sensor data fusion, decision support, web information service, e-commerce, online trading, and dataintensive smart space applications. Furthermore, the information system is being globalized due to the fast growth of the Internet. Despite the importance and wide applicability, the performance and predictability of a database system \Gamma the core component of global information systems \Gamma are relatively limited compared to the other real-time systems such as real-time operating systems. It can not be easily replicated due to the consistency problem. In addition, the database system has relatively low predictability compared to other real-time systems due to data dependence of the transaction execution, data and resource conflicts, dynamic paging and I/O, and transaction aborts and the resulting rollbacks and restarts [36]. Because of the limited perfo DB 391515 Learning Comprehensible Conceptual User Models for User Adaptive Meta Web Search In course of the OySTER project our goal is to induce conceptual user models that allow for a transparent query refinement and information filtering in the domain of Www meta--search. User models which describe a user's interest with respect to an underlying ontology allow for a manual user model editing process and also pose a well defined problem for a conceptual inductive learning task. OySTER is a research prototype that is currently being developed at the university of Osnabruck. Introduction User Modeling and Machine Learning. User models represent assumptions about a user. User modeling systems infer user models from user interaction, store user models and induce new assumptions by reasoning about the models. These models are used within the system in order to adapt to the user. Furthermore, these models shall be accessible to the user --- they should be both understandable and manually modifyable. Incorporating machine learning into this framework often leads to intertwine... IR picco99lime LIME: Linda Meets Mobility Lime is a system designed to assist in the rapid development of dependable mobile applications over both wired and ad hoc networks. Mobile agents reside on mobile hosts and all communication takes place via transiently shared tuple spaces distributed across the mobile hosts. The decoupled style of computing characterizing the Linda model is extended to the mobile environment. At the application level, both agents and hosts perceive movement as a sudden change of context. The set of tuples accessible by a particular agent residing on a given host is altered transparently in response to changes in the connectivity pattern among the mobile hosts. In this paper we present the key design concepts behind the Lime system. 1 INTRODUCTION Today's users demand ubiquitous network access independent of their physical location. This style of computation, often referred to as mobile computing, is enabled by rapid advances in the wireless communication technology. The networking scenarios enabled ... Agents aroyo02layered A Layered Approach towards Domain Authoring Support This paper presents an approach to authoring support for Web courseware based on a layered ontological paradigm. The ontology-based layers in the courseware authoring architecture serve as a basis for formal semantics and reasoning support in performing generic authoring tasks. This approach represents an extension of our knowledge classification and indexing mechanism from a previously developed system, AIMS, aimed at supporting students while completing learning tasks in a Web-based learning/training environment. We propose the addition of two vertical layers in the system architecture, Author assisting layer and Operational layer, with the role of facilitating the creation of the ontological layers (Course ontology and Domain ontology) and of the educational metadata layer. Here we focus on the domain ontology creation process, together with the support that the additional layers can provide within this process. We exemplify our method by presenting a set of generic tasks related to conceptbased domain authoring and their ontological support IR kwok01scaling Scaling Question Answering to the Web The wealth of information on the web makes it an attractive resource for seeking quick answers to simple, factual questions such as “who was the first American in space? ” or “what is the second tallest mountain in the world? ” Yet today’s most advanced web search services (e.g., Google and AskJeeves) make it surprisingly tedious to locate answers to such questions. In this paper, we extend question-answering techniques, first studied in the information retrieval literature, to the web and experimentally evaluate their performance. First we introduce MULDER, which we believe to be the first general-purpose, fully-automated question-answering system available on the web. Second, we describe MULDER’s architecture, which relies on multiple search-engine queries, natural-language parsing, and a novel voting procedure to yield reliable answers coupled with high recall. Finally, we compare MULDER’s performance to that of Google and AskJeeves on questions drawn from the TREC-8 question answering track. We find that MULDER’s recall is more than a factor of three higher than that of AskJeeves. In addition, we find that Google requires 6.6 times as much user effort to achieve the same level of recall as MULDER. IR kim00heterogeneous Heterogeneous Multimedia Database Selection on the Web Multimedia databases on the Web have two distinct characteristics, autonomy and heterogeneity. They are established and maintained independently and queries are processed depending on their own schemes. In this paper, we investigate the problem of the database selection from a number of multimedia databases dispersed on the Web. In the multimedia objects retrieval from distributed sites, it is crucial that the metaserver has the capability to find objects, globally similar to a given query object, from different multimedia databases with different local similarity measures. We propose a novel selection algorithm to determine candidate databases that contain more objects relevant to the query than other databases. The selection of databases is based on the number of relevant objects at each local database that is estimated by using a few sample objects and the histogram information. An extensive experiment on a large number of real data sets demonstrates that our proposed method perform... IR abonyi02datadriven Data-Driven Generation of Compact, Accurate, and Linguistically-Sound Fuzzy Classifiers Based on a Decision-Tree Initialization The data-driven identification of fuzzy rule-based classifiers for high-dimensional problems is addressed. A binary decision-tree-based initialization of fuzzy classifiers is proposed for the selection of the relevant features and e#ective initial partitioning of the input domains of the fuzzy system. Fuzzy classifiers have more flexible decision boundaries than decision trees (DTs) and can therefore be more parsimonious. Hence, the decision tree initialized fuzzy classifier is reduced in an iterative scheme by means of similarity-driven rule-reduction. To improve classification performance of the reduced fuzzy system, a genetic algorithm with a multi-objective criterion searching for both redundancy and accuracy is applied. The proposed approach is studied for (i) an artificial problem, (ii) the Wisconsin Breast Cancer classification Preprint submitted to Elsevier Science 19 May 2002 problem, and (iii) a summary of results is given for a set of well-known classification problems available from the Internet: Iris, Ionospehere, Glass, Pima, and Wine data. AI karadkar00evolution Evolution of the Walden's Paths Authoring Tools : Changing user skills, available infrastructure, and work practices have caused many differences in the authoring support provided by the Walden's Paths project since its conception. In this paper we trace these changes and the transition from the earlier authoring tools that supported an integrated authoring process, to the more recent tools designed to work with the Web applications that teachers have become accustomed to. 1. Introduction Hypertext has come a long way from being found only in research systems to being a part of our everyday lives in the form of the World-Wide Web (WWW or the Web). We use the Web for browsing academic information, for furthering business interests, for entertainment and a variety of other purposes. There is an immense amount of information on the Web that can be used for a variety of reasons. Web-based information can be harnessed to supplement classroom teaching for K-12 students. K-12 teachers can use Web-based information in the curriculum t... IR flank98layered A Layered Approach To Nlp-Based Information Retrieval A layered approach to information retrieval permits the inclusion of multiple search engines as well as multiple databases, with a natural language layer to convert English queries for use by the various search engines. The NLP layer incorporates morphological analysis, noun phrase syntax, and semantic expansion based on WordNet. 1 Introduction This paper describes a layered approach to information retrieval, and the natural language component that is a major element in that approach. The layered approach, packaged as Intermezzo TM , was deployed in a pre-product form at a government site. The NLP component has been installed, with a proprietary IR engine, PhotoFile, (Flank, Martin, Balogh and Rothey, 1995), (Flank, Garfield, and Norkin, 1995), at several commercial sites, including Picture Network International (PNI), Simon and Schuster, and John Deere. Intermezzo employs an abstraction layer to permit simultaneous querying of multiple databases. A user enters a query into a clien... IR yu99finding Finding the Most Similar Documents across Multiple Text Databases In this paper, we present a methodology for finding the n most similar documents across multiple text databases for any given query and for any positive integer n. This methodology consists of two steps. First, databases are ranked in a certain order. Next, documents are retrieved from the databases according to the order and in a particular way. If the databases containing the n most similar documents for a given query can be ranked ahead of other databases, the methodology will guarantee the retrieval of the n most similar documents for the query. A statistical method is provided to identify databases, each of which is estimated to contain at least one of the n most similar documents. Then, a number of strategies is presented to retrieve documents from the identified databases. Experimental results are given to illustrate the relative performance of different strategies. 1 Introduction The Internet has become a vast information source in recent years and can be considered as the w... IR japan98learning Learning to Perceive the World as Articulated: An Approach for Hierarchical Learning in Sensory-Motor Systems This paper describes how agents can learn an internal model of the world structurally by focusing on the problem of behavior-based articulation. We develop an on-line learning scheme -- the so-called mixture of recurrent neural net (RNN) experts -- in which a set of RNN modules becomes self-organized as experts on multiple levels in order to account for the different categories of sensory-motor flow which the robot experiences. Autonomous switching of activated modules in the lower level actually represents the articulation of the sensory-motor flow. In the meanwhile, a set of RNNs in the higher level competes to learn the sequences of module switching in the lower level, by which articulation at a further more abstract level can be achieved. The proposed scheme was examined through simulation experiments involving the navigation learning problem. Our dynamical systems analysis clarified the mechanism of the articulation; the possible correspondence between the articulation... AI marques01integrating Integrating Mobile Agents into Off-the-Shelf Web Servers: The M&M Approach The mobile agent paradigm provides a new approach for developing distributed systems. During the last two years, we have been working on a project that tries to overcome some of the limitations found in terms of programmability and usability of the mobile agent paradigm in real applications. In the M&M framework there are no agent platforms. Instead applications become agent-enabled by using simple JavaBeans components. In our approach the agents arrive and departure directly from the applications, interacting with them from the inside. Agents 435716 Flexible Queries to Semi-structured Datasources: the WG-log Approach A line of research is presented aimed at specifying both logical and navigational aspects of semi-structured data sources such as Web sites through the unifying notion of schema. Gracefully supporting schemata that are huge or subject to change, the WG-Log language allows for a uniform representation of queries and views, the latter expressing customized access structures to site information. A survey of related work and some directions for future research involving fuzzy query techniques are also outlined. 1 Introduction and Motivations Modern network-oriented information systems often have to deal with data that are semi-structured, i.e. lack the strict, regular, and complete structure required by traditional database management systems (see [Abi97] and [Suc97] for a survey on semi-structured data and related research). Information is semi-structured also when the structure of data varies w.r.t. time, rather than w.r.t. space: even if data is fairly well structured, such struc... IR garland00learning Learning Task Models for Collagen For an application-independent collaborative tool, a key step is to develop a detailed task model for a particular domain. This is a time consuming and dicult task, and seems to require a fairly advanced knowledge of AI representations for plans, goals, and recipes. This paper discusses some preliminary ideas for making it easier to construct and evolve task models, either through interaction with a human domain expert, through machine learning, or in a mixed-initiative system. Introduction An important trend in recent work on human-computer interaction and user modeling has been to view humancomputer interaction as a kind of collaboration. In this approach, the human user and the computer, often personied as an \agent," coordinate their actions toward achieving shared goals. Collagen is an application-independent collaboration manager based on the SharedPlan theory of task-oriented collaborative discourse (Grosz & Sidner, 1990; Rich & Sidner, 1998; Lesh, Rich, & Sidner, 199... HCI krishnamurthy01dynamic A Dynamic Replica Selection Algorithm for Tolerating Timing Faults Server replication is commonly used to improve the fault tolerance and response time of distributed services. An important problem when executing time-critical applications in a replicated environment is that of preventing timing failures by dynamically selecting the replicas that can satisfy a client's timing requirement, even when the quality of service is degraded due to replica failures and excess load on the server. In this paper, we describe the approach we have used to solve this problem in AQuA, a CORBA-based middleware that transparently replicates objects across a local area network. The approach we use estimates a replica's response time distribution based on performance measurements regularly broadcast by the replica. An online model uses these measurements to predict the probability with which a replica can prevent a timing failure for a client. A selection algorithm then uses this prediction to choose a subset of replicas that can together meet the client's timing constraints with at least the probability requested by the client. We conclude with experimental results based on our implementation. IR 162738 Using Knowledge Containers to Model a Framework for Learning Adaptation Knowledge In this paper we present a framework for learning adaptation knowledge which knowledge light approaches for case-based reasoning (CBR) systems. Knowledge light means that these approaches use already acquired knowledge inside the CBR system. Therefor we describe the sources of knowledge inside a CBR system along the different knowledge containers. After that we present our framework in terms of these knowledge containers. Further we apply our framework in a case study to one knowledge light approach for learning adaptation knowledge. After that we point on some issues which should be addressed during the design or the use of such algorithms for learning adaptation knowledge. From our point of view many of these issues should be the topic of further research. Finally we close with a short discussion and an outlook to further work. ML haritsa00realtime Real-Time Index Concurrency Control Real-time database systems are expected to rely heavily on indexes to speed up data access and thereby help more transactions meet their deadlines. Accordingly, highperformance index concurrency control (ICC) protocols are required to prevent contention for the index from becoming a bottleneck. In this paper, we develop real-time variants of a representative set of classical B-tree ICC protocols and, using a detailed simulation model, compare their performance for real-time transactions with firm deadlines. We also present and evaluate a new real-time ICC protocol called GUARD-link that augments the classical B-link protocol with a feedback-based admission control mechanism. Both point and range queries, as well as the undos of the index actions of aborted transactions are included in the scope of our study. The performance metrics used in evaluating the ICC protocols are the percentage of transactions that miss their deadlines and the fairness with respect to transaction type and size... DB dwyer99flow Flow Analysis for Verifying Specifications of Concurrent and Distributed Software This paper presents FLAVERS, a finite state verification approach that analyzes whether concurrent or sequential programs satisfy user-defined correctness properties. In contrast to other finite-state verification techniques, FLAVERS is based on algorithms with low-order polynomial bounds on the running time. FLAVERS achieves this efficiency at the cost of precision. Users, however, can improve the precision of the results by selectively and judiciously incorporating additional semantic information into the analysis problem. The FLAVERS analysis approach has been implemented for programs written in Ada. We report on an empirical study of the performance of applying the FLAVERS/Ada tool set to a collection of multi-tasking Ada programs. This study indicates that sufficient precision for proving program properties can be achieved and that the cost for such analysis grows as a low-order polynomial in the size of the program. 1 Introduction The application of distributed and concurrent pr... DB mcbreen01evaluating Evaluating Humanoid Synthetic Agents in E-Retail Applications This paper presents three experiments designed to empirically evaluate humanoid synthetic agents in electronic retail applications. Firstly, human-like agents were evaluated in a single e-retail application, a home furnishings service. The second experiment explored application dependency effects by evaluating the same human-like agents in a different e-retail application, a personalized CD service. The third experiment evaluated the effectiveness of a range of humanoid cartoon-like agents. Participants eavesdropped on spoken dialogues between a "customer" and each of the agents, which played the role of conversational sales assistants. Results showed participants expected a high level of realistic human-like verbal and nonverbal communicative behavior from the human-like agents. Overall ratings of the agents showed no significant application dependency: Two different groups of participants rated the human-like agents in similar ways in a different application. Further results showed participants have a preference for three--dimensional (3-D) rather than two--dimensional (2-D) cartoon-like agents and have a desire to interact with fully embodied agents. I. Agents zhang99towards Towards Transparent Control of Large and Complex Systems System Identi cation. Unlike with Markovian Decision Processes, some systems' output depends not only on the current state, but also the previous input/output. As a training data set for nonlinear system identi cation, the Box-Jenkins gas furnace data [BJ70] is often studied and compared. The furnace input is the gas ow rate x(t), the output y(t) is the CO 2 concentration. At least 10 candidate inputs are considered: x(t 6); x(t 5); : : : ; x(t 1); y(t 1); : : : ; y(t 4). If all of them are used, building a fuzzy controller means to solve a 10-input-1-output problem. If each input is de ned by 5 linguistic terms, this would result in a fuzzy rule system of about 10 million rules. The modelling and prediction o IR 458790 Audio Driven Facial Animation For Audio-Visual Reality In this paper, we demonstrate a morphing based automated audio driven facial animation system. Based on an incoming audio stream, a face image is animated with full lip synchronization and expression. An animation sequence using optical flow between visemes is constructed, given an incoming audio stream and still pictures of a face speaking different visemes. Rules are formulated based on coarticulation and the duration of a viseme to control the continuity in terms of shape and extent of lip opening. In addition to this new viseme-expression combinations are synthesized to be able to generate animations with new facial expressions. Finally various applications of this system are discussed in the context of creating audio-visual reality. 1. HCI dubois00augmented Augmented Reality: Which Augmentation for Which Reality? In this paper, we first present a brief review of approaches used for studying and designing Augmented Reality (AR) systems. The variety of approaches and definitions in AR requires classification. We define two intrinsic characteristics of AR systems, task focus and nature of augmentation. Based on these two characteristics, we identify four classes of AR systems. In addition our OP-a-S notation provides a complementary characterization method based on interaction. Using OP-a-S, an AR system is modeled as a set of components that communicate with each other. One crucial type of OP-a-S component is the Adapter that establishes a bridge between the real world and the virtual world. By defining a classification scheme, we aim at providing a better understanding of the paradigm of AR and at laying the foundations of future design principles according to the class of systems. Keywords Classification. Interaction characterization. INTRODUCTION One of the recent design goals in Human Com... HCI 284772 Meeting Plan Recognition Requirements for Real-Time Air-Mission Simulations In this paper, the potential synergy between instancebased pattern recognition and means-end (possible world) reasoning is explored, for supporting plan recognition in multi-aeroplane air-mission simulations. A combination of graph matching, induction, probabilistic principles and dynamic programming are applied to traces of aeroplane behaviour during flight manoeuvres. These satisfy the real-time constraints of the simulation. This enables the agents to recognise what other agents are doing and to abstract about their activity, at the instrumentation level. A means-end-reasoning model is then used to deliberate about and invoke standard operating procedures, based on recognised activity. The reasoning model constrains the recognition process by framing queries according to what a pilot would expect during the execution of the current plan(s). Results from experiments involving the dMARS procedural reasoning system and the CLARET pattern matching and induction system are described for ... Agents 455346 Accelerating Reinforcement Learning through the Discovery of Useful Subgoals An ability to adjust to changing environments and unforeseen circumstances is likely to be an important component of a successful autonomous space robot. This paper shows how to augment reinforcement learning algorithms with a method for automatically discovering certain types of subgoals online. By creating useful new subgoals while learning, the agent is able to accelerate learning on a current task and to transfer its expertise to related tasks through the reuse of its ability to attain subgoals. Subgoals are created based on commonalities across multiple paths to a solution. We cast the task of finding these commonalities as a multiple-instance learning problem and use the concept of diverse density to find solutions. We introduced this approach in [10] and here we present additional results for a simulated mobile robot task. 1 ML park98alternative Alternative Correctness Criteria for Multiversion Concurrency Control and a Locking Protocol via Freezing Concurrency control protocols based on multiversions have been used in some commercial transaction processing systems in order to provide the serializable executions of transactions. In the existing protocols, transactions are allowed to read only the most recent version of each data item in order to ensure the correct execution of transactions. However, this feature is not desirable in some advanced database systems which have more requirements such as timing or security constraints besides serializability. In this paper, we propose a new correctness criteria, called F-serializability, for multiversion concurrency control protocols. It is the extended definition of `1-serial' and relaxes the condition so that a protocol provides not only the most recent version but also the previous one to transactions, if necessary. We prove that whenever a multiversion schedule is F-serializable, the schedule is also one-copy serializable. This is the first contribution of our paper. Next, we propos... DB seymore99learning Learning Hidden Markov Model Structure for Information Extraction Statistical machine learning techniques, while well proven in fields such as speech recognition, are just beginning to be applied to the information extraction domain. We explore the use of hidden Markov models for information extraction tasks, specifically focusing on how to learn model structure from data and how to make the best use of labeled and unlabeled data. We show that a manually-constructed model that contains multiple states per extraction field outperforms a model with one state per field, and discuss strategies for learning the model structure automatically from data. We also demonstrate that the use of distantly-labeled data to set model parameters provides a significant improvement in extraction accuracy. Our models are applied to the task of extracting important fields from the headers of computer science research papers, and achieve an extraction accuracy of 92.9%. Introduction Hidden Markov modeling is a powerful statistical machine learning technique that is just ... ML 201909 Agent Aided Aircraft Maintenance Aircraft maintenance is performed bymechanics who are required, for non-standard discrepancies, to consult expert engineers for repair instructions and approval. In addition to their own experience, these engineers rely on external information sources, which are often inadequately indexed and geographically dispersed. The timely retrieval of this distributed information is essencial to the engineers' abilityto devise and recommend repair procedures in response to the mechanics ' requests. This problem domain is well suited for a multi-agent system: it consists of distributed multi-modal information whichis needed bymultiple users with diverse preferences. In this paper, we describe an implementation of such a system, using the RETSINA multi-agentarchitecture. Such an implementation reinforces the importance of multi-agent systems, and in particular the usefulness of the RETSINA infrastructure as a basis for the construction of such systems. i 1 Introduction Agent aided... Agents ghani00data Data Mining on Symbolic Knowledge Extracted from the Web Information extractors and classifiers operating on unrestricted, unstructured texts are an errorful source of large amounts of potentially useful information, especially when combined with a crawler which automatically augments the knowledge base from the world-wide web. At the same time, there is much structured information on the World Wide Web. Wrapping the web-sites which provide this kind of information provide us with a second source of information; possibly less up-to-date, but reliable as facts. We give a case study of combining information from these two kinds of sources in the context of learning facts about companies. We provide results of association rules, propositional and relational learning, which demonstrate that data-mining can help us improve our extractors, and that using information from two kinds of sources improves the reliability of data-mined rules. 1. INTRODUCTION The World Wide Web has become a significant source of information. Most of this computer-retri... IR sloman99building Building Cognitively Rich Agents Using The Sim Agent Toolkit Synthetic agents with varying degrees of intelligence and autonomy are being designed in many research laboratories. The motivations include military training simulations, games and entertainments, educational software, digital personal assistants, software agents managing Internet transactions or purely scientific curiosity. Different approaches are being explored, including, at one extreme, research on the interactions between agents, and at the other extreme research on processes within agents. The first approach focuses on forms of communication, requirements for consistent collaboration, planning of coordinated behaviours to achieve collaborative goals, extensions to logics of action and belief for multiple agents, and types of emergent phenomena when many agents interact, for instance taking routing decisions on a telecommunications network. The second approach focuses on the internal architecture of individual agents required for social interaction, collaborative behaviours, complex decision making, learning, and emergent phenomena within complex agents. Agents with complex internal structure may, for example, combine perception, motive generation, planning, plan execution, execution monitoring, and even emotional reactions. Agents 51036 Stratum Approaches to Temporal DBMS Implementation database interfaces, legacy systems. Previous approaches to implementing temporal DBMSs have assumed that a temporal DBMS must be built from scratch, employing an integrated architecture and using new temporal implementation techniques such as temporal indexes and join algorithms. However, this is a very large and time-consuming task. This paper explores approaches to implementing a temporal DBMS as a stratum on top of an existing non-temporal DBMS, rendering implementation more feasible by reusing much of the functionality of the underlying conventional DBMS. More specifically, the paper introduces three stratum meta-architectures, each with several specific architectures. Based on a new set of evaluation criteria, advantages and disadvantages of the specific architectures are identified. The paper also classifies all existing temporal DBMS implementations according to the specific architectures they employ. It is concluded that a stratum architecture is the best short, medium, and perhaps even longterm, approach to implementing a temporal DBMS. 1 DB menczer01evaluating Evaluating Topic-Driven Web Crawlers Due to limited bandwidth, storage, and computational resources, and to the dynamic nature of the Web, search engines cannot index every Web page, and even the covered portion of the Web cannot be monitored continuously for changes. Therefore it is essential to develop effective crawling strategies to prioritize the pages to be indexed. The issue is even more important for topic-specific search engines, where crawlers must make additional decisions based on the relevance of visited pages. However, it is difficult to evaluate alternative crawling strategies because relevant sets are unknown and the search space is changing. We propose three different methods to evaluate crawling strategies. We apply the proposed metrics to compare three topic-driven crawling algorithms based on similarity ranking, link analysis, and adaptive agents. IR ahamad99scalable Scalable Consistency Protocols for Distributed Services AbstractÐA common way to address scalability requirements of distributed services is to employ server replication and client caching of objects that encapsulate the service state. The performance of such a system could depend very much on the protocol implemented by the system to maintain consistency among object copies. We explore scalable consistency protocols that never require synchronization and communication between all nodes that have copies of related objects. We achieve this by developing a novel approach called local consistency (LC). LC based protocols can provide increased flexibility and efficiency by allowing nodes control over how and when they become aware of updates to cached objects. We develop two protocols for implementing strong consistency using this approach and demonstrate that they scale better than a traditional invalidation based consistency protocol along the system load and geographic distribution dimensions of scale. Index TermsÐScalable services, distributed objects, replication, caching, consistency protocols. 1 DB 35804 Query Optimization for Selections using Bitmaps Bitmaps are popular indexes for data warehouse (DW) applications and most database management systems offer them today. This paper proposes query optimization strategies for selections using bitmaps. Both continuous and discrete selection criteria are considered. Query optimization strategies are categorized into static and dynamic. Static optimization strategies discussed are the optimal design of bitmaps, and algorithms based on tree and logical reduction. The dynamic optimization discussed is the approach of inclusion and exclusion for both bit-sliced indexes and encoded bitmap indexes. 1 Introduction Bitmap indexing has become a promising technique for query processing in DWs. Variations of bitmap indexes include bit-sliced indexes [14, 3], encoded bitmap indexes (EBI) [18], bitmapped join indexes [13], range-based bitmap indexes [20], and others[16]. For query operations, such as selections, aggregates, and joins, query evaluation algorithms using bitmaps have been proposed in ... DB graef00construction Construction of Adaptive Web-Applications from Reusable Components . The Web has become a ubiquitous environment for application HCI ghani00using Using Error-Correcting Codes For Text Classification This paper explores in detail the use of Error Correcting Output Coding (ECOC) for learning text classifiers. We show that the accuracy of a Naive Bayes Classifier over text classification tasks can be significantly improved by taking advantage of the error-correcting properties of the code. We also explore the use of different kinds of codes, namely Error-Correcting Codes, Random Codes, and Domain and Data-specific codes and give experimental results for each of them. The ECOC method scales well to large data sets with a large number of classes. Experiments on a real-world data set show a reduction in classification error by up to 66% over the traditional Naive Bayes Classifier. We also compare our empirical results to semitheoretical results and find that the two closely agree. 1. Introduction Text Classification is the problem of grouping text documents into classes or categories. For the purpose of this paper, we define classification as categorizing documents in... IR puliafito00using Using Mobile Agents to implement flexible Network Management strategies Due to their intrinsic complexity, computer and communication systems require increasingly more sophisticated management starategies to be adopted in order to guarantee adequate levels of performance and reliability. The centralized paradigm adopted by the SNMP is appropriate in several network management applications, but the quick expansion of networks has posed the problem of its scalability, as well as for any other centralized model. Mobile agents represent a challenging approach to provide advanced network management functionalities, due to the possibility to easily implement a decentralized and active monitoring of the system. In this paper we discuss how to take advantage of this technology and identify some reference scenario where mobile agents represent a very promising approach. We also describe a prototype implementation based on our mobile agent platform called MAP and show how it is possible to take advantages from using the features it provides. Keywords: net... Agents hattori99socialware Socialware: Multiagent Systems for Supporting Network Communities ing with credit is permitted. To copy otherwise, to republish, to post on servers, or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from Publications Dept, ACM Inc., fax +1 (212) 869-0481, or permissions@acm.org. Community B Community A Community C Community agent A1 Community agent B1 Community agent B2 Personal unit 1 Personal unit 2 Personal unit 5 Personal unit 6 Personal unit 3 Personal unit 7 Personal unit 4 User 5 to join User 4 to leave Figure 1: A general architecture of socialware as a multiagent system Socialware as Multiagent Systems There are several characteristics specific to network communities, which make a multiagent architecture attractive to use. The first characteristic is that the participants of a network community are widely distributed and the number of potential participants is large. Hence, no solid, centralized, or monolithic system would be adequate: A distributed system would be required, in which perso... Agents dragoni00updating Updating Mental States from Communication . In order to perform effective communication agents must be able to foresee the effects of their utterances on the addressee's mental state. In this paper we investigate on the update of the mental state of an hearer agent as a consequence of the utterance performed by a speaker agent. Given an agent communication language with a STRIPSlike semantics, we propose a set of criteria that allow to bind the speaker's mental state to its uttering of a certain sentence. On the basis of these criteria, we give an abductive procedure that the hearer can adopt to partially recognize the speaker's mental state that led to a specific utterance. This procedure can be adopted by the hearer to update its own mental state and its image of the speaker's mental state. 1 Introduction In multi-agent systems, communication is necessary for the agents to cooperate and coordinate their activities or simply to avoid interfering with one another. If agents are not designed with embedded pre-compiled... Agents cruz99preserving Preserving Contextual Navigation in Hypermedia Querying Navigating query results is a highly volatile task, usually requiring much effort from the users, while not providing a firm reference point for further queries or refinement. Many users accessing traditional search engines are confronted with lengthy pages of hypertext links, through which relevant information must be found. Accessing each link can bring a user closer or farther from the information they seek; in either case, the context between the query results and the current web page can be lost. Delaunay MM addresses these issues by creating virtual documents through which all query results are displayed in a meaningful and organized fashion; and that reference the originating web page, thus providing semantic browsing and hypermedia navigation without loss of context. Keywords: Digital Library Querying, Hypermedia Querying, Context Preservation Introduction With an increase in the number of users daily, the World Wide Web has become one of the most indispensable technologie... IR 124833 Supporting Dialogue Inferencing in Conversational Case-Based Reasoning . Dialogue inferencing is the knowledge-intensive process of inferring aspects of a user's problem from its partial description. Conversational case-based reasoning (CCBR) systems, which interactively and incrementally elicit a user's problem description, suffer from poor retrieval efficiency (i.e., they prompt the user with questions that the user has already implicitly answered) unless they perform dialogue inferencing. The standard method for dialogue inferencing in CCBR systems requires library designers to supply explicit inferencing rules. This approach is problematic (e.g., maintenance is difficult). We introduce an alternative approach in which the CCBR system guides the library designer in building a domain model. This model and the partial problem description are then given to a query retrieval system (PARKA-DB) to infer any implied answers during a conversation. In an initial empirical evaluation in the NaCoDAE CCBR tool, our approach improved retrieval efficiency without sa... AI reitmayr01wearable A Wearable 3D Augmented Reality Workspace This poster describes our work to build a wearable Augmented Reality system that supports true stereoscopic 3D graphics. Through a pen and pad interface well known 2D user interfaces can be presented to the user, whereas the tracking of the pen allows to use direct interaction with virtual objects. The system is assembled from off-the-shelf hardware components and serves as a basic test bed for user interface experiments related to collaboration between stationary and mobile AR users. 1. HCI 4579 An Architecture and Object Model for Distributed Object-Oriented Real-Time Databases The confluence of computers, communications, and databases is quickly creating a distributed database where many applications require real-time access to both temporally accurate and multimedia data. This is particularly true in military and intelligence applications, but these required features are needed in many commercial applications as well. We are developing a distributed database, called BeeHive, which could offer features along different types of requirements: real-time, fault-tolerance, security, and quality-of service for audio and video. Support of these features and potential trade-offs between them could provide a significant improvement in performance and functionality over current distributed database and object management systems. In this paper, we present a high level design for BeeHive architecture and sketch the design of the BeeHive Object Model (BOM) which extends object-oriented data models by incorporating time and other features into objects, resulting in a high... DB luck01conceptual A Conceptual Framework for Agent Definition and Development The use of agents of many different kinds in a variety of fields of computer science and artificial intelligence is increasing rapidly and is due, in part, to their wide applicability. The richness of the agent metaphor that leads to many different uses of the term is, however, both a strength and a weakness: its strength lies in the fact that it can be applied in very many different ways in many situations for different purposes; the weakness is that the term agent is now used so frequently that there is no commonly accepted notion of what it is that constitutes an agent. This paper addresses this issue by applying formal methods to provide a defining framework for agent systems. The Z specification language is used to provide an accessible and unified formal account of agent systems, allowing us to escape from the terminological chaos that surrounds agents. In particular, the framework precisely and unambiguously provides meanings for common concepts and terms, enables alternative models of particular classes of system to be described within it, and provides a foundation for subsequent development of increasingly more refined concepts. Agents 288424 Principles for a Usability-Oriented Pattern Language High-level usability principles (e.g. "Familiarity") are difficult to apply to specific projects, and style guides providing more detailed instructions are often misinterpreted and inaccessible. An approach to usability based on design patterns enables designers to learn how certain recurring problems can be solved according to high-level principles. This paper summarises a review of the desirable properties advocated by five popular style guides, and discusses how this list has been modified to provide an underlying philosophy which is appropriate for a usability-oriented pattern language. A sample pattern which exemplifies this philosophy, involving iteration through selectable objects, is described. KEYWORDS Usability engineering, Design techniques, Style guides. 1. Introduction There has been considerable discussion about how to reconcile the gaps between software engineering (SE) and human-computer interaction (HCI). One of the primary ways to smoothly integrate the disciplines ... HCI inoue01controlling Controlling Speculative Computation in Multi-Agent Environments In this paper, we propose a multi-agent system which performs speculative computation under incomplete communication environments. In a master-slave style multi-agent system with speculative computation, a master agent asks queries to slave agents in problem solving, and proceeds computation with default answers when answers from slave agents are delayed. We rst provide a semantics for speculative computation using default logic. Then, in the proposed system, we use the consequence- nding procedure SOL written in the Java language to perform data-driven deductive reasoning. The use of a consequence- nding procedure is convenient for updating agents' beliefs according to situation changes in the world. In our system, slave agents can change their answers frequently, yet a master agent can avoid duplicate computation. As long as actual answers from slave agents do not con- ict with any previously encountered situation, the obtained conclusions are never recomputed. We applied the proposed system to the meeting-room reservation problem to see the usefulness of the framework. 1 Agents 845 Learning to Catch: Applying Nearest Neighbor Algorithms to Dynamic Control Tasks Steven L. Salzberg 1 and David W. Aha 2 1 Introduction Dynamic control problems are the subject of much research in machine learning (e.g., Selfridge, Sutton, & Barto, 1985; Sammut, 1990; Sutton, 1990). Some of these studies investigated the applicability of various k-nearest neighbor methods (Dasarathy, 1990) to solve these tasks by modifying control strategies based on previously gained experience (e.g., Connell & Utgoff, 1987; Atkeson, 1989; Moore, 1990; 1991). However, these previous studies did not highlight the fact that small changes in the design of these algorithms drastically alter their learning behavior. This paper describes a preliminary study that investigates this issue in the context of a difficult dynamic control task: learning to catch a ball moving in a three-dimensional space, an important problem in robotics research (Geng et al., 1991). Our thesis in this paper is that agents can improve substantially at physical tasks by storing experiences without explicitly... ML 332262 Ensembles of Classifiers Based on Approximate Reducts . A problem of improving rough set based expert systems by modifying a notion of reduct is discussed. A notion of approximate reduct is introduced, as well as some proposals of quality measure for such a reduct. A complete classifying system based on approximate reducts is presented and discussed. It is proved that a problem of nding optimal set of classifying agents based on approximate reducts is NP-hard; a genetic algorithm is used to nd the suboptimal set. Experimental results show, that the classifying system is eective and relatively fast. 1 Introduction Rough set expert systems base on the notion of reduct [11], [12], a minimal subset of attributes which is sucient to discern between objects with dierent decision values. A set of short reducts can be used to generate rules [2]. A problem of short reducts generation is NP-hard, but an approximate algorithm (like the genetic one described in [13], [8] and implemented successfully { see [10]) can be used to obtain redu... AI baek01cost Cost Effective Mobile Agent Planning for Distributed Information Retrieval The number of agents' and the execution time are two significant performance factors' in mobile agent planning. Fewer agents' cause lower network traffic and consume less bandwidth. Regardless of the number of agents' used, the execution time for a task must be kept minimal, which means that use of the minimal number of agents' must not impact on the execution time unfavorably. As the population of the mobile agent application domain grows', the importance of these two factors' also increases. Agents nottelmann01learning Learning probabilistic Datalog rules for information classification and transformation Probabilistic Datalog is a combination of classical Datalog (i.e., function-free Horn clause predicate logic) with probability theory. Therefore, probabilistic weights may be attached to both facts and rules. But it is often impossible to assign exact rule weights or even to construct the rules themselves. Instead of specifying them manually, learning algorithms can be used to learn both rules and weights. In practice, these algorithms are very slow because they need a large example set and have to test a high number of rules. We apply a number of extensions to these algorithms in order to improve efficiency. Several applications demonstrate the power of learning probabilistic Datalog rules, showing that learning rules is suitable for low dimensional problems (e.g., schema mapping) but inappropriate for higher dimensions like e.g. in text classification. IR hazelhurst98binary Binary Decision Diagram Representations Of Firewall And Router Access Lists Network firewalls and routers can use a rule database to decide which packets will be allowed from one network onto another. By filtering packets the firewalls and routers can improve security and performance -- by excluding packets which may pose a security risk to a network or are not relevant to it. However, as the size of the rule list increases, it becomes difficult to maintain and validate the rules, and the cost of rule lookup may add significantly to latency. Ordered binary decision diagrams (BDDs) -- a compact method of representing and manipulating boolean expressions -- are a potential method of representing the rules. This paper explores how BDDs can be used to develop methods that help analysis of rules to validate them and changes to them, to improve performance, and facilitate hardware support. 1 Introduction The growth of network and internet communication creates several challenges for network design. Two important issues are security and performance. As the volume o... DB 274436 Collision Avoidance and Resolution Multiple Access for Multichannel Wireless Networks The CARMA-NTG protocol is presented and analyzed. CARMA-NTG dynamically divides the channel into cycles of variable length; each cycle consists of a contention period and a group-transmission period. During the contention period, a station with one or more packets to send competes for the right to be added to the group of stations allowed to transmit data without collisions; this is done using a collision resolution splitting algorithm based on a request-to-send/clear-to-send (RTS/CTS) message exchange with non-persistent carrier sensing. CARMA-NTG ensures that one station is added to the group transmission period if one or more stations send requests to be added in the previous contention period. The group-transmission period is a variable-length train of packets, which are transmitted by stations that have been added to the group by successfully completing an RTS/CTS message exchange in previous contention periods. As long as a station maintains its position in the group, it is able to transmit data packets without collision. An upper bound is derived for the average costs of obtaining the first success in the splitting algorithm. This bound is then applied to the computation of the average channel utilization in a fully connected network with a large number of stations. These results indicate that collision resolution is a powerful mechanism in combination with floor acquisition and group allocation multiple access. 1 HCI chang98data Data Resource Selection in Distributed Visual Information Systems With the advances in multimedia databases and the popularization of the Internet, it is now possible to access large image and video repositories distributed throughout the world. One of the challenging problems in such an access is how the information in the respective databases can be summarized to enable an intelligent selection of relevant database sites based on visual queries. This paper presents an approach to solve this problem based on image content-based indexing of a metadatabase at a query distribution server. The metadatabase records a summary of the visual content of the images in each database through image templates and statistical features characterizing the similarity distributions of the images. The selection of the databases is done by searching the metadatabase using a ranking algorithm that uses query similarity to a template and the features of the databases associated with the template. Two selection approaches, termed mean-based and histogram-based approaches, ... IR 300852 Centroid-Based Document Classification: Analysis & Experimental Results . In this paper we present a simple linear-time centroid-based document classification algorithm, that despite its simplicity and robust performance, has not been extensively studied and analyzed. Our experiments show that this centroid-based classifier consistently and substantially outperforms other algorithms such as Naive Bayesian, k-nearest-neighbors, and C4.5, on a wide range of datasets. Our analysis shows that the similarity measure used by the centroidbased scheme allows it to classify a new document based on how closely its behavior matches the behavior of the documents belonging to different classes. This matching allows it to dynamically adjust for classes with different densities and accounts for dependencies between the terms in the different classes. 1 Introduction We have seen a tremendous growth in the volume of online text documents available on the Internet, digital libraries, news sources, and company-wide intranets. It has been forecasted that these docu... IR jantke94nonstandard Nonstandard Concepts of Similarity in Case-Based Reasoning Introduction The present paper is aimed at propagating new concepts of similarity more flexible and expressive than those underlying most case-based reasoning approaches today. So, it mainly deals with criticizing approaches in use, with motivating and introducing new notions and notations, and with first steps towards future applications. The investigations at hand originate from the author's work in learning theory. In exploring the relationship between inductive learning and case-based learning within a quite formal setting (cf. [Jan92b]), it turned out that both areas almost coincide, if sufficiently flexible similarity concepts are taken into acount. This provides some formal arguments for the necessity of non-symmetric similarity measures. Encouraged by these first results, the author tried to investigate more structured learning problems from the view point of case-based reasoning. It turned out that an appropriate handling requires formalisms allowing similarity concep ML warshaw99rulebased Rule-Based Query Optimization, Revisited We present an overview and initial performance assessment of a rule-based query optimizer written in VenusDB. VenusDB is an active-database rule language embedded in C++. Following the developments in extensible database query optimizers, first in rule-based form followed by optimizers written as object-oriented programs, the VenusDB optimizer avails the advantages of both. To date, development of rule-based query optimizers have included the definition and implementation of custom rule languages. Thus, extensibility required detailed understanding and often further development of the underlying search mechanism of the rule system. Objectoriented query optimizers appear to have achieved their goals with respect to a clear organization and encapsulation of an optimizer's elements. They do not, however, provide for the concise, declarative expression of domain specific heuristics. Our experience demonstrates that a rule-based query optimizer developed in VenusDB can be well structured, ... DB tzouramanis00multiversion Multiversion Linear Quadtree for Spatio-Temporal Data Research in spatio-temporal databases has largely focused on extensions of access methods for the proper handling of time changing spatial information. In this paper, we present the Multiversion Linear Quadtree (MVLQ), a spatio-temporal access method based on Multiversion B-trees (MVBT) [2], embedding ideas from Linear Region Quadtrees [4]. More specifically, instead of storing independent numerical data having a different transaction-time each, for every consecutive image we store a group of codewords that share the same transaction-time, whereas each codeword represents a spatial subregion. Thus, the new structure may be used as an index mechanism for storing and accessing evolving raster images. We also conducted a thorough experimentation using sequences of real and synthetic raster images. In particular, we examined the time performance of temporal window queries, and provide results for a variety of parameter settings. DB nie01mining Mining Source Coverage Statistics for Data Integration Recent work in data integration has shown the importance of statistical information about the coverage and overlap of sources for efficient query processing. Despite this recognition there are no effective approaches for learning the needed statistics. The key challenge in learning such statistics is keeping the number of needed statistics low enough to have the storage and learning costs manageable. Naive approaches can become infeasible very quickly. In this paper we present a set of connected techniques that estimate the coverage and overlap statistics while keeping the needed statistics tightly under control. Our approach uses a hierarchical classification of the queries, and threshold based variants of familiar data mining techniques to dynamically decide the level of resolution at which to learn the statistics. We describe the details of our method, and present preliminary experimental results showing the feasibility of the approach. 1. DB 458848 Bringing Robustness to End-User Programming In some cases, end-user programming allows the design of stand-alone applications. But none of the existing approaches is concerned by safety aspects of programming. Heavy techniques exist to develop safe applications, particularly in non-interactive domains. They involve software engineering techniques, and sometimes, formal methods. All these techniques are very far from end-users. Our idea is to let this part to experts, and to connect end-user programming onto this safe conventional development. Starting from an existing functional core, we built an interactive end-user programming environment called GenBuild, which allows designing interactive stand-alone applications. GenBuild allows the verification of some properties that are a first step towards the development of safe end-user programming. 1. HCI sakakibara94learning Learning Languages by Collecting Cases and Tuning Parameters . We investigate the problem of case-based learning of formal languages. Case-based reasoning and learning is a currently booming area of artificial intelligence. The formal framework for case-based learning of languages has recently been developed by [JL93] in an inductive inference manner. In this paper, we first show that any indexed class of recursive languages in which finiteness is decidable is case-based representable, but many classes of languages including the class of all regular languages are not case-based learnable with a fixed universal similarity measure, even if both positive and negative examples are presented. Next we consider a framework of case-based learning where the learning algorithm is allowed to learn similarity measures, too. To avoid trivial encoding tricks, we carefully examine to what extent the similarity measure is going to be learned. Then by allowing only to learn a few parameters in the similarity measures, we show that any indexed class of recursive ... ML scassellati98building Building Behaviors Developmentally: A New Formalism This paper advocates a developmental approach to building complex interactive behaviors for robotic systems. A developmental methodology is advantageous because it provides a structured decomposition of complex tasks, because it facilitates learning, and because it allows for a gradual increase in task complexity. The developmental approach provides a structured means both of dividing a task among research units, as well as a metric for evaluating the progress of the task. Initial work with developmental modeling has also hinted that these skill decompositions may make the overall task easier to accomplish through the re-use of knowledge gained from developmental precursors. We report here on two projects of building behaviors developmentally for a humanoid robot. In the first project, the robot learned to reach to a visual target by following a developmental progression similar to those observed in infants. The second project outlines a proposal for building social skills developmenta... AI grimm01systems Systems Directions for Pervasive Computing Pervasive computing, with its focus on users and their tasks rather than on computing devices and technology, provides an attractive vision for the future of computing. But, while hardware and networking infrastructure to realize this vision are becoming a reality, precious few applications run in this infrastructure. We believe that this lack of applications stems largely from the fact that it is currently too hard to design, build, and deploy applications in the pervasive computing space. In this paper, we argue that existing approaches to distributed computing are flawed along three axes when applied to pervasive computing; we sketch out alternatives that are better suited for this space. First, application data and functionality need to be kept separate, so that they can evolve gracefully in a global computing infrastructure. Second, applications need to be able to acquire any resource they need at any time, so that they can continuously provide their services in a highly dynamic environment. Third, pervasive computing requires a common system platform, allowing applications to be run across the range of devices and to be automatically distributed and installed. 1. Agents azuma99challenge The Challenge of Making Augmented Reality Work Outdoors this paper, along with others we have yet to imagine. Acknowledgments HCI 455692 SI-Designer: an Integration Framework for E-Commerce Electronic commerce lets people purchase goods and exchange information on business transactions on-line. Therefore one of the main challenges for the designers of the e-commerce infrastructures is the information sharing, retrieving data located in different sources thus obtaining an integrated view to overcome any contradiction or redundancy. Virtual Catalogs synthesize this approach as they are conceived as instruments to dynamically retrieve information from multiple catalogs and present product data in a unified manner, without directly storing product data from catalogs. In this paper we propose SI-Designer, a support tool for the integration of data from structured and semi-structured data sources, developed within the MOMIS (Mediator environment for Multiple Information Sources) project. 1 DB roy99coastal Coastal Navigation - Mobile Robot Navigation with Uncertainty in Dynamic Environments Ships often use the coasts of continents for navigation in the absence of better tools such as GPS, since being close to land allows sailors to determine with high accuracy where they are. Similarly for mobile robots, in many environments global and accurate localization is not always feasible. Environments can lack features, and dynamic obstacles such as people can confuse and block sensors. In this paper, we demonstrate a technique for generating trajectories that take into account both the information content of the environment, and the density of the people in the environment. These trajectories reduce the average positional certainty as the robot moves, reducing the likelihood the robot will become lost at any point. Our method was successfully implemented and used by the mobile robot Minerva, a museum tourguide robot, for a 2 week period in the Smithsonian National Museum of American History. 1 Introduction One essential component of any operational mobile robot system is the ab... AI 364207 The KRAFT Architecture for Knowledge Fusion and Transformation This paper describes the KRAFT architecture which supports the fusion of knowledge from multiple, distributed, heterogeneous sources. The architecture uses constraints as a common knowledge interchange format, expressed against a common ontology. Knowledge held in local sources can be tranformed into the common constraint language, and fused with knowledge from other sources. The fused knowledge is then used to solve some problem or deliver some information to a user. Problem-solving in KRAFT typically exploits pre-existing constraint solvers. KRAFT uses an open and flexible agent architecture in which knowledge sources, knowledge fusing entities, and users are all represented by independent KRAFT agents, communicating using a messaging protocol. Facilitator agents perform matchmaking and brokerage services between the various kinds of agent. KRAFT is being applied to an example application in the domain of network data services design. 1 Introduction and Motivation Most ... Agents 20543 Nonmonotonic Inheritance in Object-Oriented Deductive Database Languages Deductive object-oriented frameworks integrate logic rules and inheritance. There, specific problems arise: Due to the combination of deduction and inheritance, (a) deduction can take place depending on inherited facts, thus raising indirect conflicts, and (b) also the class hierarchy and-membership is subject to deduction. From this point of view, we investigate the application of the extension semantics of Default Logic to deductive object-oriented database languages. By restricting the problem to Horn programs and a special type of defaults tailored to the semantics of inheritance, a forwardchaining construction of extensions is possible. This construction is compared with a solution as implemented in the F-Logic system Florid which is based on a combination of classical deductive fixpoints and an inheritance-trigger mechanism. The paper is a condensed version of [MK98]. 1 Introduction In deductive object-oriented database languages, a class hierarchy and nonmonotonic inheritance i... DB baldoni00intentionguided Intention-guided Web Sites: A New Perspective on Adaptation Recent years witnessed a rapid growth of multimedia technologies for offering information and services on the internet. One of the many problems that are to be faced in this context is the great variety of possible users and the consequent need to adapt both the presentation of information and the interaction to the specific user's characteristics. Agents 138121 An Algorithm to Evaluate Quantified Boolean Formulae and its Experimental Evaluation The high computational complexity of advanced reasoning tasks such as reasoning about knowledge and planning calls for efficient and reliable algorithms for reasoning problems harder than NP. In this paper we propose Evaluate, an algorithm for evaluating Quantified Boolean Formulae, a language that extends propositional logic in a way such that many advanced forms of propositional reasoning, e.g., circumscription, can be easily formulated as evaluation of a QBF. Algorithms for evaluation of QBFs are suitable for the experimental analysis on a wide range of complexity classes, a property not easily found in other formalisms. Evaluate is based on a generalization of the Davis-Putnam procedure for SAT, and is guaranteed to work in polynomial space. Before presenting the algorithm, we discuss several abstract properties of QBFs that we singled out to make it more efficient. We also discuss various options that were investigated about heuristics and data structures, and report the main res... AI kopp00planning Planning and Motion Control in Lifelike Gesture: A Refined Approach In this paper an operational model for the automatic generation of lifelike gestures of an anthropomorphic virtual agent is described. The biologically motivated approach to controlling the movements of a highly articulated figure provides a transformation of spatiotemporal gesture specifications into an analog representation of the movement from which the animations are directly rendered. To this end, knowledge-based computer animation techniques are combined with appropriate methods for trajectory formation and articulated figure animation. 1. Introduction The inclusion of nonverbal modalities into the communicative behaviors of virtual agents has moved into focus of human-computer interface researchers. Humans are more likely to consider computer-generated figures lifelike when appropriate nonverbal behaviors are displayed in addition to speech. This enables the evocation of social communicative attributions to the artificial agent, which are supposed to be advantageous for a natu... Agents rauber01integrating Integrating Automatic Genre Analysis into Digital Libraries With the number and types of documents in digital library systems increasing, tools for automatically organizing and presenting the content have to be found. While many approaches focus on topic-based organization and structuring, hardly any system incorporates automatic structural analysis and representation. Yet, genre information (unconsciously) forms one of the most distinguishing features in conventional libraries and in information searches. In this paper we present an approach to automatically analyze the structure of documents and to integrate this information into an automatically created content-based organization. In the resulting visualization, documents on similar topics, yet representing dierent genres, are depicted as books in diering colors. This representation supports users intuitively in locating relevant information presented in a relevant form. Keywords Genre Analysis, Self-Organizing Map (SOM), SOMLib, Document Clustering, Visualization, Metaphor Graphics 1. IR 374383 Localisation using Automatically Selected Landmarks from Panoramic Images The use of visual landmarks for robot localisation is a promising field. It is apparent that the success of localisation by visual landmarks depends on the landmarks chosen. Good landmarks are those which remain reliable over time and through changes in position and orientation. This paper describes a system which learns places by automatically selecting landmarks from panoramic images and uses them for localisation tasks. An adaption of the biologically inspired Turn Back and Look behaviour is used to evaluate potential landmarks. Normalised correlation is used to overcome the a#ects of changes in illumination in the environment. Results from real robot experiments are reported, showing successful localisation from up to one meter away from the learnt position. 1 Introduction Visual localisation is one of the key problems in making successful autonomous robots. Vision as a sensor is the richest source of information about a mobile agent's environment and as such con... Agents ragsdale00adaptation Adaptation Techniques for Intrusion Detection and Intrusion Response Systems This paper examines techniques for providing adaptation in intrusion detection and intrusion response systems. As attacks on computer systems are becoming increasingly numerous and sophisticated, there is a growing need for intrusion detection and response systems to dynamically adapt to better detect and respond to attacks. The Adaptive Hierarchical Agentbased Intrusion Detection System (AHA! IDS) provides detection adaptation by adjusting the amount of system resources devoted to the task of detecting intrusive activities. This is accomplished by dynamically invoking new combinations of lower level detection agents in response to changing circumstances and by adjusting the confidence associated with these lower-level agents. The Adaptive Agentbased Intrusion Response System (AAIRS) provides response adaptation by weighting those responses that have been successful in the past over those techniques that have not been as successful. As a result, the more successful responses are used... AI bertino01measuring Measuring the Structural Similarity among XML Documents and DTDs Sources of XML documents are proliferating on the Web and documents are more and more frequently exchanged among sources. At the same time, there is an increasing need of exploiting database tools to manage this kind of data. An important novelty of XML is that information on document structures is available on the Web together with the document contents. This information can be used to improve document handling and to achieve more effective and efficient searches on documents. However, in such an heterogeneous environment as the Web, it is not reasonable to assume that XML documents that enter a source always conform to a predefined DTD present in the source. IR 314563 Scheduling Queries on Tape-resident Data Advances in storage technology have made near-line tertiary storage a viable alternative for database and data warehouse systems. Tertiary storage systems are employed in cases where secondary storage can not satisfy the data handling requirements or tertiary storage is more cost effective option. Tertiary storage devices have traditionally been used as archival storage. The new application domains require on-demand retrieval of data from these devices. This paper investigates issues in optimizing I/O time for a query whose data resides on automated tertiary storage containing multiple storage devices. We model the problem as a limited storage parallel two-machine flow-shop scheduling problem with additional constraints. Given a query, we establish an upper bound on the number of storage devices for an optimal I/O schedule and provide experimental proof for it. For queries that access small amounts of data from multiple media, we derive an optimal schedule analytically. For q... DB knowles99evolutionary Evolutionary Approaches to Off-Line Routing in Backbone Communications Networks Off-line routing in backbone communications networks is an important combinatorial optimisation problem. It has three main uses: first, off-line routing provides reference benchmark results for dynamic (on-line) routing strategies. Second, and more interestingly, off-line routing is becoming more and more investigated and employed in its own right as a way of quickly finding significantly improved routings for live networks which can then be imposed on the network to offer a net improvement in quality of service. Third, it can be used in networks where bandwidth may be booked in advance. In this paper we introduce and investigate a number of heuristic techniques applicable to the routing problem for use in stochastic, iterative search. Results are presented which indicate that these heuristics significantly improve the search for solutions, particularly when on-line performance is considered. We also investigate how computation time can be further reduced by the use of delta-evaluation... ML 132577 Optimizing Object Queries Using an Effective Calculus Object-oriented databases (OODBs) provide powerful data abstractions and modeling facilities but they generally lack a suitable framework for query processing and optimization. One of the key factors for OODB systems to successfully compete with relational systems as well as to meet the performance requirements of many non-traditional applications is the development of an effective query optimizer. We propose an effective framework with a solid theoretical basis for optimizing OODB query languages. Our calculus, called the monoid comprehension calculus, captures most features of ODMG OQL and is a good basis for expressing various optimization algorithms concisely. This paper concentrates on query unnesting, an optimization that, even though improves performance considerably, is not treated properly (if at all) by most OODB systems. Our framework generalizes many unnesting techniques proposed recently in the literature and is capable of removing any form of query nesting using a very si... DB rao97unified A Unified View of Plans as Recipes Plans as recipes or abstract structures, as well as plans as mental attitudes that guide an agent in its planning process has been enthusiastically embraced by both philosophers and AI practitioners. They play a central role in a class of rational agents, called Belief-Desire-Intention (BDI) agents. This dual view of plans can not only be used for efficient planning, but can also be used for recognizing the plans of other agents, coordinating one's actions and achieving joint intentions with other members of a larger collective or team, and finally recognizing the collective plans and intentions of other teams. In this paper, we start with a simple notion of execution plans and discuss its operational semantics. We progressively extend this notion of plans to recognition plans, joint execution plans, and joint recognition plans. The primary contribution of this paper is in providing an integrated view of plans that facilitate individual an collective planning and recognition. 1 Int... Agents arlein99making Making LDAP Active with the LTAP Gateway: Case Study in Providing Telecom Integration and Enhanced Services LDAP (Lightweight Directory Access Protocol) directories are being rapidly deployed on the Web. They are currently used to store data like white pages information, user profiles, and network device descriptions. These directories offer a number of advantages over current database technology in that they provide better support for heterogeneity and scalability. However, they lack some basic database functionality (e.g., triggers, transactions) that is crucial for Directory Enabled Networking (DEN) tasks like provisioning network services, allocating resources, reporting, managing end-to-end security, and offering mobile users customized features that follow them. In order to address these limitations while keeping the simplicity and performance features of LDAP directories, unbundled and portable solutions are needed. In this paper we discuss LDAP limitations we faced while building an LDAP meta-directory that integrates data from legacy telecom systems, and how LTAP (Lightweight Trigger Access Process), a portable gateway that adds active functionality to LDAP directories, overcomes these limitations. DB aksoy98textural Textural Features for Image Database Retrieval This paper presents two feature extraction methods and two decision methods to retrieve images having some section in them that is like the user input image. The features used are variances of gray level co-occurrences and line-angle-ratio statistics constituted by a 2-D histogram of angles between two intersecting lines and ratio of mean gray levels inside and outside the regions spanned by those angles. The decision method involves associating with any pair of images either the class “relevant” or “irrelevant”. A Gaussian classifier and nearest neighbor classifier are used. A protocol that translates a frame throughout every image to automatically define for any pair of images whether they are in the relevance class or the irrelevance class is discussed. Experiments on a database of 300 gray scale images with 9,600 groundtruth image pairs showed that the classifier assigned 80 % of the image pairs we were sure were relevant, to the relevance class correctly. The actual retrieval accuracy is greater than this lower bound of 80%. DB carkacioglu02learning Learning Similarity Space In this study, we suggest a method to adapt an image retrieval system into a configurable one. Basically, original feature space of a content-based retrieval system is nonlinearly transformed into a new space, where the distance between the feature vectors is adjusted by learning. The transformation is realized by Artificial Neural Network architecture. A cost function is defined for learning and optimized by simulated annealing method. Experiments are done on the texture image retrieval system, which use Gabor Filter features. The results indicate that configured image retrieval system is significantly better than the original system. 1. AI bruno01stholes STHoles: A Multidimensional Workload-Aware Histogram Attributes of a relation are not typically independent. Multidimensional histograms can be an effective tool for accurate multiattribute query selectivity estimation. In this paper, we introduce STHoles,a\workload-aware" histogram that allows bucket nesting to capture data regions with reasonably uniform tuple density. STHoles histograms are built without examining the data sets, but rather by just analyzing query results. Buckets are allocated where needed the most as indicated by the workload, which leads to accurate query selectivity estimations. Our extensive experiments demonstrate that STHoles histograms consistently produce good selectivity estimates across synthetic and real-world data sets and across query workloads, and, in many cases, outperform the best multidimensional histogram techniques that require access to and processing of the full data sets during histogram construction. 1. DB phan01new A New TWIST on Mobile Computing: Two-Way Interactive Session Transfer The ubiquitous use of computer resources for daily productivity is a goal that presently remains unrealised. We believe that the convergence of desktop and mobile applications into a seamless computing experience will provide a strong motivation for future "anytime, anywhere computing." In this paper we describe this convergence as the capability to perform the handoff of application sessions across heterogeneous platforms using the network as a conduit. In addition to discussing the architecture and protocols to facilitate this capability, in this paper we also provide a taxonomy for describing a variety of different session handoff schemes. In particular, we have identified an important Two-Way Interactive Session Transfer (TWIST) behaviour for communication between heterogeneous clients and servers. To demonstrate our concepts, we have implemented the handoff capability with TWIST semantics into a real-world application that serves as a teaching tool for radiology clinicians. From experimental data we will show that the handoff mechanism incurs little delay to transfer large dataladen sessions. 1 HCI accot97beyond Beyond Fitts' Law: Models for Trajectory-Based HCI Tasks Trajectory-based interactions, such as navigating through nested-menus, drawing curves, and moving in 3D worlds, are becoming common tasks in modern computer interfaces. Users' performances in these tasks cannot be successfully modeled with Fitts' law as it has been applied to pointing tasks. Therefore we explore the possible existence of robust regularities in trajectory-based tasks. We used "steering through tunnels" as our experimental paradigm to represent such tasks, and found that a simple "steering law" indeed exists. The paper presents the motivation, analysis, a series of four experiments, and the applications of the steering law. HCI mchugh98indexing Indexing Semistructured Data This paper describes techniques for building and exploiting indexes on semistructured data: data that may not have a fixed schema and that may be irregular or incomplete. We first present a general framework for indexing values in the presence of automatic type coercion. Then based on Lore, a DBMS for semistructured data, we introduce four types of indexes and illustrate how they are used during query processing. Our techniques and indexing structures are fully implemented and integrated into the Lore prototype. 1 Introduction We call data that is irregular or that exhibits type and structural heterogeneity semistructured, since it may not conform to a rigid, predefined schema. Such data arises frequently on the Web, or when integrating information from heterogeneous sources. In general, semistructured data can be neither stored nor queried in relational or object-oriented database management systems easily and efficiently. We are developing Lore 1 , a database management system d... DB ordonez00discovering Discovering Interesting Association Rules in Medical Data We are presently exploring the idea of discovering association rules in medical data. There are several technical aspects which make this problem challenging. In our case medical data sets are small, but have high dimensionality. Information content is rich: there exist numerical, categorical, time and even image attributes. Data records are generally noisy. We explain how to map medical data to a transaction format suitable for mining rules. The combinatorial nature of association rules matches our needs, but current algorithms are unsuitable for our purpose. We thereby introduce an improved algorithm to discover association rules in medical data which incorporates several important constraints. Some interesting results obtained by our program are discussed and we explain how the program parameters were set. We believe many of the problems we come across are likely to appear in other domains. 1 Introduction Data Mining is an active research area. One of the most popular approaches t... AI 471747 MARIAN: Flexible Interoperability for Federated Digital Libraries Federated digital libraries are composed of distributed autonomous (heterogeneous) information services but provide users with a transparent, integrated view of collected information – respecting different information sources’ autonomy. In this paper we discuss a federated system for the Networked Digital Library of Theses and Dissertations (NDLTD), an international consortium of universities, libraries, and other supporting institutions focused on electronic theses and dissertations (ETDs). The NDLTD has so far allowed its members considerable autonomy, though agreements are developing on metadata standards and on support of the Open Archives initiative that eventually will promote greater homogeneity. At present, federation requires dealing flexibly with differences among systems, ontologies, and data formats. Our solution involves adapting MARIAN, an objectoriented digital library retrieval system developed with support by NLM and NSF, to serve as mediation middleware for the federated NDLTD collection. Components of the solution include: 1) the use of several harvesting techniques; 2) an architecture based on object-oriented ontologies of search modules and metadata; 3) diversity within the harvested data joined to a single collection view for the user; and 4) an integrated framework for addressing such questions as data quality, information compression, and flexible search. The system can handle very large dynamic collections. An adaptable relationship between the collection view and harvested data facilitates adding new sites to the federation and adapting to changes in existing sites. MARIAN’s modular architecture and powerful and flexible data model work together to build an effective integrated solution within a simple uniform framework. We present both the general design of the system and operational details of a preliminary federated collection involving several thousand ETDs in four different formats and two languages from USA and Europe. IR starner98visual Visual Contextual Awareness in Wearable Computing Small, body-mounted video cameras enable a different style of wearable computing interface. As processing power increases, a wearable computer can spend more time observing its user to provide serendipitous information, manage interruptions and tasks, and predict future needs without being directly commanded by the user. This paper introduces an assistant for playing the real-space game Patrol. This assistant tracks the wearer's location and current task through computer vision techniques and without off-body infrastructure. In addition, this paper continues augmented reality research, started in 1995, for binding virtual data to physical locations. 1. Introduction For most computer systems, even virtual reality systems, sensing techniques are a means of getting input directly from the user. However, wearable computers offer a unique opportunity to re-direct sensing technology towards recovering more general user context. Wearable computers have the potential to "see" as the user sees... HCI elio00task Task Models, Intentions, and Agent Conversation Policies . It is possible to define conversation policies, such as communication or dialogue protocols, that are based strictly on what messages and, respectively, what performatives may follow each other. While such an approach has many practical applications, such protocols support only "local coherence" in a conversation. Lengthy message exchanges require some infrastructure to lend them "global coherence." Recognition of agent intentions about the joint task is essential for this global coherence, but there are further mechanisms needed to ensure that both local and global coherence are jointly maintained. This paper presents a general yet practical approach to designing, managing, and engineering agents that can do simple run-time intention recognition without creating complex multi-state protocols. In this approach we promote developing abstract task models and designing conversation policies in terms of such models. An implemented agent assistant based on these ideas is brie... Agents devedzic01knowledge Knowledge Modeling - State of the Art A major characteristic of developments in the broad field of Artificial Intelligence (AI) during the 1990s has been an increasing integration of AI with other disciplines. A number of other computer science fields and technologies have been used in developing intelligent systems, starting from traditional information systems and databases, to modern distributed systems and the Internet. This paper surveys knowledge modeling techniques that have received most attention in recent years among developers of intelligent systems, AI practitioners and researchers. The techniques are described from two perspectives, theoretical and practical. Hence the first part of the paper presents major theoretical and architectural concepts, design approaches, and research issues. The second part discusses several practical systems, applications, and ongoing projects that use and implement the techniques described in the first part. Finally, the paper briefly covers some of the most recent results in the fields of intelligent manufacturing systems, intelligent tutoring systems, and ontologies. 2 1. Agents 450405 Coverage Problems in Wireless Ad-hoc Sensor Networks Wireless ad-hoc sensor networks have recently emerged as a premier research topic. They have great long term economic potential, ability to transform our lives, and pose many new system building challenges. Sensor networks also pose a number of new conceptual and optimization problems. Some, such as location, deployment, and tracking, are fundamental issues, in that many applications rely on them for needed information. In this paper, we address one of the fundamental problems, namely coverage. Coverage in general, answers the questions about quality of service (surveillance) that can be provided by a particular sensor network. We first define the coverage problem from several points of view including deterministic, statistical, worst and best case, and present examples in each domain. By combining computational geometry and graph theoretic techniques, specifically the Voronoi diagram and graph search algorithms, we establish the main highlight of the paper - optimal polynomial time worst and average case algorithm for coverage calculation. We also present comprehensive experimental results and discuss future research directions related to coverage in sensor networks. I. HCI 111340 Run-time Detection in Parallel and Distributed Systems: An Application to Safety-Critical Applications As systems are becoming more complex, there is increasing interest in their runtime analysis, understanding their dynamic behavior and possibly controling it as well. This paper describes complex distributed and parallel applications that use run-time analyses to attain scalability improvements with respect to the amount and complexity of the data transmitted, transformed, and shared among different application components. Such improvements are derived from using database techniques when manipulating data streams. Namely, by imposing a relational model on a data stream, filters and constraints on the stream may be expressed in the form of database queries evaluated against the data events comprising the stream. Streams may then be filtered using runtime optimization techniques derived from query optimization methods. This paper also presents a tool, termed Cnet, which offers (1) means for the dynamic creation of queries and their application to distributed data streams, (2) permits the... DB vanhoof98specialising Specialising The Other Way Around In this paper, we present a program transformation based on bottom up evaluation of logic programs. We explain that using this technique, programs can be "specialised" w.r.t. a set of unit clauses instead of a query. Moreover, examples indicate that excellent specialisation can be obtained when this bottom up transformation is combined with a more traditional top down approach, resulting in conceptually cleaner techniques requiring a less complicated control than one overall approach. 1 DB kaburlasos00learning Learning and Decision-Making in the Framework of Fuzzy Lattices A novel theoretical framework is delineated for supervised and unsupervised learning. It is called framework of fuzzy lattices, or FLframework for short, and it suggests mathematically sound tools for dealing separately and/or jointly with disparate types of data including vectors of numbers, fuzzy sets, symbols, etc. Specific schemes are proposed for clustering and classification having the capacity to deal with both missing and don't care data values; the schemes in question can be implemented as neural networks. The proposed learning schemes are employed here for pattern recognition on seven data sets including benchmark data sets, and the results are compared with those ones by various learning techniques from the literature. Finally, aiming at a mutual cross-fertilization, the FL-framework is associated with established theories for learning and/or decision-making including probability theory, fuzzy set theory, Bayesian decision-making, theory of evidence, and adaptive resonance t... ML tresp98description A Description Logic for Vague Knowledge This work introduces the concept language ALCFM which is an extension of ALC to many-valued logics. ALCFM allows to express vague concepts, e.g. more or less enlarged or very small. To realize this extension to many-valued logics, the classical notions of satisfiability and subsumption had to be modified appropriately. For example, ALCFM -concepts are no longer either satisfiable or unsatisfiable, but they are satisfiable to a certain degree. The main contribution of this paper is a sound and complete method for computing the degree of subsumption between two ALCFM -concepts. 1 Introduction This work takes its motivation from the occurrence of vague concept descriptions in different application areas. Often, application--inherent information is characterized by a very high degree of vagueness. Appropriate information systems must be able to process this kind of data. So far, there are no systems that really solve the corresponding problems due to the lack of powerful basic methods. A... AI 531549 Towards Robust Teams with Many Agents Agents in deployed multi-agent systems monitor other agents to coordinate and collaborate robustly. However, as the number of agents monitored is scaled up, two key challenges arise: (i) the number of monitoring hypotheses to be considered can grow exponentially in the number of agents; and (ii) agents become physically and logically unconnected (unobservable) to their peers. This paper examines these challenges in teams of cooperating agents, focusing on a monitoring task that is of particular importance to robust teamwork: detecting disagreements among team-members. We present YOYO, a highly scalable disagreement-detection algorithm which guarantees sound detection in time linear in the number of agents despite the exponential number of hypotheses. In addition, we present new upper bounds about the number of agents that must be monitored in a team to guarantee disagreement detection. Both YOYO and the new bounds are explored analytically and empirically in thousands of monitoring problems, scaled to thousands of agents. Agents weiss98intelligent Intelligent Telecommunication Technologies anage telecommunication networks. Building such applications involved acquiring valuable telecommunication knowledge from human experts and then applying this knowledge, typically by embedding it in an expert system. This knowledge acquisition process is so time-consuming that it is referred to as the "knowledge acquisition bottleneck". Data mining techniques are now being applied to industrial applications to break this bottleneck, by replacing the manual knowledge acquisition process with automated knowledge discovery. Telecommunication networks, which routinely generate tremendous amounts of data, are ideal candidates for data mining [1]. This section will describe expert system and data mining technologies and how they are evolving to solve complex industrial problems. 1.1 Expert Systems Expert systems are programs which represent and apply factual knowledge of specific areas of expertise to solve problems [2]. Expert systems have been applied extensively within ML marsella00interactive Interactive Pedagogical Drama This paper describes an agent-based approach to realizing interactive pedagogical drama. Characters choose their actions autonomously, while director and cinematographer agents manage the action and its presentation in order to maintain story structure, achieve pedagogical goals, and present the dynamic story to as to achieve the best dramatic effect. Artistic standards must be maintained while permitting substantial variability in story scenario. To achieve these objectives, scripted dialog is deconstructed into elements that are portrayed by agents with emotion models. Learners influence how the drama unfolds by controlling the intentions of one or more characters, who then behave in accordance with those intentions. Interactions between characters create opportunities to move the story in pedagogically useful directions, which the automated director exploits. This approach is realized in the multimedia title Carmen's Bright IDEAS, an interactive health intervention designed to impro... Agents dubois99classification Classification Space for Augmented Surgery, an Augmented Reality Case Study One of the recent design goals in Human Computer Interaction has been to extend the sensorymotor capabilities of computer systems to combine the real and the virtual in order to assist the user in his environment. Such systems are called Augmented Reality (AR). Although AR systems are becoming more prevalent we still do not have a clear understanding of this interaction paradigm. In this paper we propose OPAS as a generic framework for classifying existing AR systems. Computer Assisted Medical Interventions (CAMI), for which the added value of AR has been demonstrated by experience, are discussed in light of OPAS. We illustrate OPAS using our system, CASPER (Computer ASsisted PERicardial puncture), a CAMI system which assists in surgical procedures (pericardial punctures). KEYWORDS: Augmented Surgery, CAMI, Augmented Reality, Classification Space 1. INTRODUCTION The term "Augmented Reality" (AR) appears in the literature usually in conjunction with the term "Virtual Reality" (VR). Th... HCI moukas96amalthaea Amalthaea: Information Discovery and Filtering using a Multiagent Evolving Ecosystem Agents are semi-intelligent programs that assist the user in performing repetitive and time-consuming tasks. Information discovery and information filtering are a suitable domain for applying agent technology. Ideas drawn from the field of autonomous agents and artificial life are combined in the creation of an evolving ecosystem composed of competing and cooperating agents. A co-evolution model of information filtering agents that adapt to the various user's interests and information discovery agents that monitor and adapt to the various on-line information sources is analyzed. Results from a number of experiments are presented and discussed. Keywords: Agents, Information Filtering, Evolution, World-Wide-Web 1 Introduction The exponential increase of computer systems that are interconnected in on-line networks has resulted in a corresponding exponential increase in the amount of information available on-line. This information is distributed among heterogeneous sources and is... IR 187693 Leveraging Mediator Cost Models with Heterogeneous Data Sources : Distributed systems require declarative access to diverse data sources of information. One approach to solving this heterogeneous distributed database problem is based on mediator architectures. In these architectures, mediators accept queries from users, process them with respect to wrappers, and return answers. Wrapper provide access to underlying data sources. To efficiently process queries, the mediator must optimize the plan used for processing the query. In classical databases, cost-estimate based query optimization is an effective method for optimization. In a heterogeneous distributed databases, cost-estimate based query optimization is difficult to achieve because the underlying data sources do not export cost information. This paper describes a new method that permits the wrapper programmer to export cost estimates (cost estimate formulas and statistics). For the wrapper programmer to describe all cost estimates may be impossible due to lack of information or burdensome due... DB 43511 The Logic Programming Paradigm DB hammer99information The Information Integration Wizard (IWiz) Project data. To illustrate the need for integration of heterogeneous data sources, suppose we have a scenario where consumers want to purchase computer-related products from one of the many e-commerce sites. However, before making the purchase they would like to gather all the relevant, available information in order to help them in their decision making process. For example, consumers may want to access product information on available desktops, laptops, software, and other accessories and check availability and pricing information. In addition, users may also want to access other online sources for related background information such as consumer reports, press releases etc. This situation is depicted in Figure 1. Typically, each source uses different tools and data modeling techniques to create and manage their data. This means, the same concept, for example, the entity software, may be described by a different term and different set of attributes in different sources (e.g DB gallwitz98erlangen The Erlangen Spoken Dialogue System EVAR: A State-of-the-Art Information Retrieval System In this paper, we present an overview of the spoken dialogue system EVAR that was developed at the University of Erlangen. In January 1994, it became accessible over telephone line and could answer inquiries in the German language about German InterCity train connections. It has since been continuously improved and extended, including some unique features, such as the processing of out--of--vocabulary words and a flexible dialogue strategy that adapts to the quality of the recognition of the user input. In fact, several different versions of the system have emerged, i.e. a subway information system, train and flight information systems in different languages, and an integrated multilingual and multifunctional system which covers German and 3 additional languages in parallel. Current research focuses on the introduction of stochastic models into the semantic analysis, on the direct integration of prosodic information into the word recognition process, on the detection of user emotion, a... AI gellersen00adding Adding Some Smartness to Devices and Everyday Things In mobile computing, context-awareness indicates the ability of a system to obtain and use information on aspects of the system environment. To implement contextawareness, mobile system components have to be augmented with the ability to capture aspects of their environment. Recent work has mostly considered locationawareness, and hence augmentation of mobile artifacts with locality. In this paper we discuss augmentation of mobile artifacts with diverse sets of sensors and perception techniques for awareness of context beyond location. We report experience from two projects, one on augmentation of mobile phones with awareness technologies, and the other on embedding of awareness technology in everyday non-digital artifacts. HCI tandler01software Software Infrastructure for Ubiquitous Computing Environments: Supporting Synchronous Collaboration with Heterogeneous Devices In ubiquitous computing environments, multiple users work with a wide range of different devices. In many cases, users interact and collaborate using multiple heterogeneous devices at the same time. The configuration of the devices should be able to change frequently due to a highly dynamic, flexible and mobile nature of new work practices. This produces new requirements for the architecture of an appropriate software infrastructure. In this paper, an architecture designed to meet these requirements is proposed. To test its applicability, this architecture was used as the basis for the implementation of BEACH, the software infrastructure of i-LAND (the ubiquitous computing environment at GMD-IPSI). It provides the functionality for synchronous cooperation and interaction with roomware components, i.e. room elements with integrated information technology. In conclusion, our experiences with the current implementation are presented. HCI santos99dynamic Dynamic User Model Construction with Bayesian Networks for Intelligent Information Queries The complexity of current software applications is overwhelming users. The need exists for intelligent interface agents to address the problems of increasing taskload that is overwhelming the human user. Interface agents could help alleviate user taskload by extracting and analyzing relevant information, and providing information abstractions of that information, and providing timely, beneficial assistance to users. Central to providing assistance to a user is the issue of correctly determining the user's intent. The Clavin project is to build an intelligent natural language query information management system. Clavin must maintain a dynamic user model of the relevant concepts in the user inquiries as they relate to the information sources. The primary goal of Clavin is to autonomously react to changes in user intent as well as the information sources, by dynamically constructing the appropriate queries relative to the changes identified. In this paper, we discuss the... Agents 533789 RF*IPF: A Weighting Scheme for Multimedia Information Retrieval Region-based approach has become a popular research trend in the field of multimedia database retrieval. In this paper, we present the Region Frequency and Inverse Picture Frequency (RF*IPF) weighting, a measure developed to unify region-based multimedia retrieval systems with textbased information retrieval systems. The weighting measure gives the highest weight to regions that occur often in a small number of images in the database. These regions are considered discriminators. With this weighting measure, we can blend image retrieval techniques with TF*IDFbased text retrieval techniques for large-scale Web applications. The RF*IPF weighting has been implemented as a part of our experimental SIMPLIcity image retrieval system and tested on a database of about 200,000 general-purpose images. Experiments have shown that this technique is effective in discriminating images of different semantics. Additionally, the overall similarity approach enables a simple querying interface for multimedia information retrieval systems. 1 IR puliafito00map MAP: Design and Implementation of a Mobile Agents Platform The recent development of telecommunication networks has contributed to the success of applications such as information retrieval and electronic commerce, as well as all the services that take advantage of communication in distributed systems. In this area, the emerging technology of mobile agents aroused considerable interest. Mobile agents are applications that can move through the network for carrying out a given task on behalf of the user. In this work we present a platform (called MAP (Mobile Agents Platform)) for the development and the management of mobile agents. The language used both for developing the platform and for carrying out the agents is Java. The platform gives the user all the basic tools needed for creating some applications based on the use of agents. It enables us to create, run, suspend, resume, deactivate, reactivate local agents, to stop their execution, to make them communicate each other and migrate. Keywords: mobile agents, distributed computing, Java, net... Agents bettini01klava Klava: a Java Framework for Distributed and Mobile Applications Highly distributed networks have now become a common infrastructure for a new kind of wide-area distributed applications whose key design principle is network awareness, namely the ability of dealing with dynamic changes of the network environment. Network-aware computing has called for new programming languages that exploit the mobility paradigm as a basic interaction mechanism. In this paper we present the architecture of Klava, an experimental Java framework for distributed applications and code mobility. We explain how Klava implements code mobility by relying on Java and show a few distributed applications that exploit mobile code and are programmed in Klava. Keywords: Code Mobility, Distributed Applications, Network Awareness, Language and Middleware Implementation, Tuple Spaces, Java. 1 Agents 27645 Optimizing Large Join Queries in Mediation Systems In data integration systems, queries posed to a mediator need to be translated into a sequence of queries to the underlying data sources. In a heterogeneous environment, with sources of diverse and limited query capabilities, not all the translations are feasible. In this paper, we study the problem of finding feasible and efficient query plans for mediator systems. We consider conjunctive queries on mediators and model the source capabilities through attribute-binding adornments. We use a simple cost model that focuses on the major costs in mediation systems, those involved with sending queries to sources and getting answers back. Under this metric, we develop two algorithms for source query sequencing -- one based on a simple greedy strategy and another based on a partitioning scheme. The first algorithm produces optimal plans in some scenarios, and we show a linear bound on its worst case performance when it misses optimal plans. The second algorithm generates optimal plans in more scenarios, while having no bound on the margin by which it misses the optimal plans. We also report on the results of the experiments that study the performance of the two algorithms. DB prodromidis99comparative A Comparative Evaluation of Meta-Learning Strategies over Large and Distributed Data Sets There has been considerable interest recently in various approaches to scaling up machine learning systems to large and distributed data sets. We have been studying approaches based upon the parallel application of multiple learning programs at distributed sites, followed by a meta-learning stage to combine the multiple models in a principled fashion. In this paper, we empirically determine the "best" data partitioning scheme for a selected data set to compose "appropriatelysized " subsets and we evaluate and compare three di#erent strategies, Voting, Stacking and Stacking with Correspondence Analysis (SCANN) for combining classification models trained over these subsets. We seek to find ways to e#ciently scale up to large data sets while maintaining or improving predictive performance measured by the error rate, a cost model, and the TP-FP spread. Keywords: classification, multiple models, meta-learning, stacking, voting, correspondence analysis, data partitioning Email address of co... ML chirkova00linearly Linearly Bounded Reformulations of Conjunctive Databases (Extended Abstract) Database reformulation is the process of rewriting the data and rules of a deductive database in a functionally equivalent manner. DB 210 A Bayesian Computer Vision System for Modeling Human Interactions AbstractÐWe describe a real-time computer vision and machine learning system for modeling and recognizing human behaviors in a visual surveillance task [1]. The system is particularly concerned with detecting when interactions between people occur and classifying the type of interaction. Examples of interesting interaction behaviors include following another person, altering one's path to meet another, and so forth. Our system combines top-down with bottom-up information in a closed feedback loop, with both components employing a statistical Bayesian approach [2]. We propose and compare two different state-based learning architectures, namely, HMMs and CHMMs for modeling behaviors and interactions. The CHMM model is shown to work much more efficiently and accurately. Finally, to deal with the problem of limited training data, a synthetic ªAlife-styleº training system is used to develop flexible prior models for recognizing human interactions. We demonstrate the ability to use these a priori models to accurately classify real human behaviors and interactions with no additional tuning or training. Index TermsÐVisual surveillance, people detection, tracking, human behavior recognition, Hidden Markov Models. 1 ML fjeld99exploring Exploring Brick-Based Camera Control Introduction BUILD-IT is a planning tool based on computer vision technology, with a capacity for complex planning and composition tasks (Rauterberg et al. 1997). The system enables users, grouped around a table, to interact in a virtual scene, using real bricks to select and manipulate objects in the scene (Fig. 1, left). A plan view of the scene is projected onto the table. A perspective view of the scene, called side view, is projected on the wall. The plan view contains a storage space with originals, allowing users to select new objects. Object selection is done by putting a brick at the object position. Once selected, objects can be positioned, rotated and fixed by simple brick manipulation (Fig. 1, right). They are de-selected, and stay put, when the brick is covered or removed. Objects brought back to the storage space are deleted. Figure 1: BUILD-IT offers a plan view for combined action and perception, and a