BSVM

Chih-Wei Hsu and Chih-Jen Lin

BSVM 2.09 released on December 15, 2018.

The two multi-class implementations included after BSVM 2.01 are two of the five methods compared in the following paper: A comparison on methods for multi-class support vector machines . (However, there is one difference: In the paper kernel caches stored numbers in double precision but in this release cached values are in single precision)

Introduction

BSVM solves support vector machines (SVM) for the solution of large classification and regression problems. It includes the following methods

One vs. One multi-class classification using a bound-constrained formulation
Multi-class classification by solving a single optimization problem (again, a bounded formulation). See Section 3 of our comparison paper.
Multi-class classification using Crammer and Singer's formulation . See Section 4 of our comparison paper.
Regression using a bound-constrained formulation
Multi-class classification using Crammer and Singer's formulation with squared hinge (L2) loss

The current implementation borrows the structure of libsvm. Similar options are also adopted. For the bound-constrained formulation for classification and regression, BSVM uses a decomposition method. BSVM uses a simple working set selection which leads to faster convergences for difficult cases. The use of a special implementation of the opmization solver TRON allows BSVM to stably identify bounded variables.

Download BSVM

The current release (Version 2.09, December 2018) of BSVM can be obtained by downloading the zip file that contains the software. The README file contains instructions on how to install the software. Please e-mail us if you have problems to download the file.

Note that BSVM is provided "as is" without express or implied warranty. This software can be freely used for research purpose. Use for commercial purposes is expressly prohibited without contacting the authors.

Additional Information

For additional information on BSVM, please see the following two papers

C.-W. Hsu and C.-J. Lin. A simple decomposition method for support vector machines , Machine Learning 46(2002), 291-314.
C.-W. Hsu and C.-J. Lin. A comparison on methods for multi-class support vector machines , IEEE Transactions on Neural Networks, 13(2002), 415-425.

For information about how BSVM handles libear SVM, see

K.-M. Chung, W.-C. Kao, T. Sun, and C.-J. Lin. Decomposition Methods for Linear Support Vector Machines

However, if you would like to use linear kernel, we recommend LIBLINEAR instead.

For information about multi-class implementations, see

J. Weston and C. Watkins, Multi-Class Support Vector Machines, Technical report CSD-TR-98-04, Royal Holloway, 1998
K. Crammer and Y. Singer, On the algorithmic implementation of multiclass kernel-based vector machines. Technical report, School of Computer Science and Engineering, Hebrew University, 2001
C.-P. Lee, and C.-J. Lin. A Study on L2-Loss (Squared Hinge-Loss) Multi-Class SVM. Technical report, 2012.

If you have any problems using BSVM, we are happy to provide help. Please send comments and suggestions to Chih-Jen Lin.

Acknowledgments: The authors thank Chih-Chung Chang for many helpful disussions and comments. Part of the software implementation also benifited from his help. BSVM 2.06 was prepared by Rong-En Fan. BSVM 2.07, 2.08 were prepared by Ching-Pei Lee.

Update log