The PSVM software contains a classification, a regression, and a feature selection mode and is based on an efficient SMO optimization technique. The software can directly be applied to dyadic (matrix) data sets or it can be used with kernels like standard SVM software. PSVM minimizes a scale-invariant capacity measure under a new set of constraints. As a result, and in contrast to standard SVMs, the kernel function does not have to be positive definite, the data matrices have not to be square, e.g. the software already implements the indefinite sin-kernel which is a non-Mercer kernel. Another important feature of the software is that is allows for n-fold cross validation and hyperparameter selection. For classification tasks it offers the determination of the significance level and ROC data. In summary the basic features of the software are
• WINDOWS and UNIX compatible • no dependencies to other software • command line interface • MATLAB interface • n-fold cross validation • hyperparameter selection • relational data • non-Mercer kernels • significance testing • computation of Receiver-Oprator-Characteristic (ROC) curves
- Changes to previous version:
Initial Announcement on mloss.org.
- BibTeX Entry: Download
- Corresponding Paper BibTeX Entry: Download
- URL: Project Homepage
- Supported Operating Systems: Platform Independent
- Data Formats: Ascii
- Tags: Svm, Classification, Regression, Feature Selection, Feature Weighting, Feature Ranking, Kernel Learning, Hyperparameter Selection, Negative Definite, Scale Invariant, Non Mercer
- Archive: download here
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.