CSL: A Comprehensive Sparse Learning Package
Danian Zheng (zhengdanian@cn.fujitus.com)
Kaizhu Huang (K.Huang@bris.ac.uk)
Version 1.0
Last update: 2009.01
Should you have any comments or suggestions, please send email to Danian Zheng or Kaizhu Huang.
Please refer to the toolbox as
Danian Zheng, Kaizhu Huang, CSL: A comprehensive Sparse Learning Package, Dept. of Engineering
Mathematics, University of Bristol, 2009.
Descriptions: Six sparse algorithms are implemented in this package. They are
1) SVM
2) L1-norm SVM
a) Linear Programming SVM
b) SC-SVM
3) L0-norm SVM
4) RVM (Relevance Vector Machine)
5) SPC (Sparse Probit Classifier)
Algorithms, codes comments, and reference can be seen in the following.
Contents:
1. SVM (Support Vector Machine)
SVC (support vector classification) is optimized by a fast SMO algorithm: (c++_dll\fsmo.cpp Æ fsmo.dll).
train_svm.m SVM training function
test_svm.m SVM test function
example_svm.m
draw_svm.m
Run “example_svm(k)” to show SVM training results on 2D examples, where k=1~ 4.
exp_svm.m Run “exp_svm(k)” to do 10-fold cross-validation experiments on benchmark datasets,
where k=1~13.