This package is an implementation of a linear RankSVM solver with non-convex regularization. This is the code that has been used for numerical simulation in the paper
Laporte, L., Flamary, R., Canu, S., Déjean, S., Mothe, J., "Nonconvex Regularizations for Feature Selection in Ranking With Sparse SVM", Neural Networks and Learning Systems, IEEE Transactions on, Vol. 25, N. 6, pp 1118-1130, 2014.
We provide a general solver for squared hinge loss RankSVM with following regularization terms : - l1 norm - lp norm with p<1 - log sum penalty. - MCP
This toolbox is in Matlab/Octave and should run on both software. The algorithm used for solving the optimization problem is a Difference of Convex approach as described in here and the algorithm used for solving the sub-problem is a Forward-Backward Splitting algorithm from the FISTA paper. The toolbox also contains code from the paper FSMrank as provided by the authors.
- Changes to previous version:
Initial Announcement on mloss.org.
- BibTeX Entry: Download
- Corresponding Paper BibTeX Entry: Download
- URL: Project Homepage
- Supported Operating Systems: Agnostic
- Data Formats: Matlab
- Tags: Matlab, Mkl, Classification, Feature Selection, Linear Svm, Convex Optimization, Gradient Based Learning, Ranking, Machine Learning, Optimization, Data Mining, Supervised Learning, Lasso, Sparsity, Regularization, Algorithm, Discriminant Analysis, Linear Classifier, L1 Minimization, L1 Norm, Gradient Based Optimization, Non Convex
- Archive: download here
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.