MSVMpack is dedicated to multi-class support vector machines (M-SVM), which can handle more than two classes without relying on decomposition methods, and implements the four M-SVM models from the literature: Weston and Watkins M-SVM, Crammer and Singer M-SVM, Lee, Lin and Wahba M-SVM, and the M-SVM2 of Guermeur and Monfrini.
MSVMpack provides the first parallel implementation of M-SVMs and use vectorized kernel functions depending on the data type for faster computations.
The software is based on two command-line tools, but also provides a full library (C API) to include its functionalities in other programs and a Matlab interface. Templates are provided to ease the addition of custom kernels.
MSVMpack now includes a web server, which offers a platform-independent GUI through a set of simple web pages. This means that training and testing can be easily controlled from a Windows computer (or even a tablet or smartphone) with network access to a Linux host running MSVMpack.
- Changes to previous version:
- Added Matlab interface.
- BibTeX Entry: Download
- Corresponding Paper BibTeX Entry: Download
- URL: Project Homepage
- JMLR MLOSS PaperURL: JMLR-MLOSS Paper Homepage
- Supported Operating Systems: Linux, Macosx, Platform Independent On The Client Side
- Data Formats: Ascii, Matlab
- Tags: Kernel, Svm, Support Vector Machines, Parallel, Msvm, Multiclass Support Vector Machine, Jmlr, Server
- Archive: download here
Other available revisons
Version Changelog Date 1.5
- Windows binaries are now included (by Emmanuel Didiot)
- MSVMpack can now be compiled on Windows (by Emmanuel Didiot)
- Fixed polynomial kernel
- Minor bug fixes
July 3, 2014, 16:02:49 1.4
- Added parallelized k-Fold cross validation ('-cv k' option in trainmsvm)
- Cache size now defaults to the maximum amount of memory
- Minor bug fixes for the Matlab interface
August 30, 2013, 10:40:35 1.3
- Added Matlab interface.
April 23, 2013, 10:44:37 1.2
- Added MSVMpack Server and Web interface
- Now becomes platform-independent at the client side
- Can import files in LibSVM data format and raw matrix format
- Model is saved periodically during training
February 7, 2013, 13:25:44 1.1
- Possibility to choose different values of C for different classes (useful for unbalanced data sets)
- Sets the number of working threads at runtime (through the command-line option '-t')
- Builds on Mac OS X
- Small changes in the model files
- Several bug fixes
October 25, 2011, 14:24:21 1.0
Bibtex entry of the corresponding JMLR paper.
July 10, 2011, 12:09:12
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.