Generalized Conditional Gradient (GCG)



GCG is an open source Matlab solver for gauge (norm) regularized problems, that are commonly used in sparse coding and compressive sensing.  Examples include matrix completion, dictionary learning, and structured sparse estimation. 

The code includes the implementations used for all experiments in [1].


Latest: GCG version 2.0 

Old stuff: GCG version 1.0 

This version 1.0 implements the boosting algorithm in [2]. Note the loss function for matrix completion needs to be customized in order to efficiently compute f(U V') where U and V have a small number of columns.  We provide a helper function here.


GCG is licensed under Mozilla Public License version 1.1. The authors are not responsible for any implications from the use of the software.


Xinhua ZhangYaoliang YuDale Schuurmans



Yaoliang Yu, Xinhua Zhang, Dale Schuurmans

Generalized Conditional Gradient for Sparse Estimation

Journal of Machine Learning Research (JMLR)

Under review, 2014.  [PDF]



Xinhua Zhang, Yaoliang Yu, Dale Schuurmans

Accelerated Training for Matrix-norm Regularization: A Boosting Approach

Advances in Neural Information Processing Systems (NIPS), 2012.  [PDF]