PSVM Download

Potential - Support Vector Machine

We offer a first implementation of the new P-SVM algorithm, a software for support vector classification, regression and feature extraction.

PSVM Documentation - Introduction to P-SVM and software features

PSVM Manual - Detailed description of all included software components and sample files

PSVM SMO Technical Report - Description of the algorithm used within the software

Software License - This software is published under the GNU General Public License

Download, compile and tryout the software version 1.31(beta) (2008-05-25) with samples and documentation files:

psvm.tar.gz - compressed files as tar-file

psvm.zip - same files in zip format

New features are:

1.31: small bugfix for classification error output of command line version.

1.3: Projection of data onto smallest rectangular containing the trainingdata is now optional (MATLAB: psvmpredict, psvmputnp; standalone: -normlimit)

1.3: Balanced errors are now consistenly supported by psvm standalone version. (option -be)

Version 1.31 from 2007-02-13: psvm.tar.gz - psvm.zip

Version 1.3 from 2007-11-19: psvm13.tar.gz - psvm13.zip

Version 1.2 from 2007-02-13: psvm12.tar.gz - psvm12.zip

Version 1.1 from 2006-06-21: psvm11.tar.gz - psvm11.zip

Version 1.0 from 2005-09-27: psvm10.zip

Last update on 2007-05-25. Please report bugs and errors to Tilman Knebel. If you use this software in your application, please cite:

Sepp Hochreiter, Klaus Obermayer; Support Vector Machines for Dyadic Data, Neural Computation, 18, 1472-1510, 2006.

External Links:

A free windows unzip software can be found here.

To take advantage of all functions of the software, GNU-Plot is needed. Click on ftp-server or a mirror and for Windows select a W32 version, and for Unix and Linux based system select a tar.gz file.

One additional dataset ARCENE.zip is referenced in the sample section of the software.

The SVM implementation libsvm is also referenced in the software, but not required.