JNI-Pegasos
Description
The software provides a java interface for the PEGASOS algorithm, which is implemented in C++ by Shai Shalev-Shwartz.

As in SVM-Light and Pegasos, GADGET SVM takes the data file in the svmlight format.
The model file, which is weight vector in the primal, is also written in sparse vector format.

An option has been added to write predicted output into a file, given by predicFile argument. Also, two test java files has been provided in the top directory named JNIPegasosTrainTest.java and JNIPegasosClassifyTest.java
Software
Available for download here
References
  • Chase Hensel and Haimonti Dutta, "GADGET SVM: a Gossip-bAseD sub-GradiEnT SVM Solver", International Conference on Machine Learning (ICML), Numerical Mathematics in Machine Learning Workshop, Montreal, Quebec, 2009.
  • Shai Shalev-Shwartz, Yoram Singer and Nathan Srebro, ``Pegasos: Primal Estimated sub-GrAdient SOlver for SVM", ICML 2007.
Related Projects
  • GADGET SVM