
 Description:
SCD is a C++ implementation of the stochastic coordinate descent algorithm proposed in
* Shai ShalevShwartz and Ambuj Tewari, Stochastic methods for l1 regularized loss minimization. Submitted to Journal of Machine Learning Research
which, in turn, is a modification of the original stochastic coordinate algorithm proposed in
* Shai ShalevShwartz and Ambuj Tewari, Stochastic methods for l1 regularized loss minimization. Proceedings of the 26th International Conference on Machine Learning, pages 929936, 2009.
It can be used for l1regularized loss minimization for both classification and regression problems.
Currently supported loss functions are the logistic loss and the squared loss. SCD is designed to run fast even for large highdimensional datasets and can exploit the sparsity in the examples.
 Changes to previous version:
Instead of keeping a vector of length 2*d as in the previous version, now the algorithm only maintain a vector of length d, where d is the number of features. This slightly reduces both the code length and runtime.
 BibTeX Entry: Download
 Corresponding Paper BibTeX Entry: Download
 URL: Project Homepage
 Supported Operating Systems: Agnostic
 Data Formats: Ascii
 Tags: Coordinate Descent, L1 Regularization, Large Datasets
 Archive: download here
Other available revisons

Version Changelog Date 2.1 Fixed some I/O bugs. Lines that ended with whitespace were not read correctly in the previous version.
December 3, 2009, 22:21:45 2.0 Instead of keeping a vector of length 2*d as in the previous version, now the algorithm only maintain a vector of length d, where d is the number of features. This slightly reduces both the code length and runtime.
November 27, 2009, 04:03:39 1.0 Initial Announcement on mloss.org.
August 5, 2009, 00:53:40
Comments
No one has posted any comments yet. Perhaps you'd like to be the first?
Leave a comment
You must be logged in to post comments.