- SparseSVMTrainer takes training data and builds a predictive model in binary format.
- SparseSVMPredictor uses the model to predict the class of each sample in a test data set.
- SVMModelPrinter displays readable information about the model.
The SparseSVMTrainer and SparseSVMPredictor functions are designed for input that is in sparse format; that is, each table row represents an attribute and each sample (observation) consists of many attributes. These functions are suitable for tasks like text classification, whose high number of attributes (many unique words) might exceed the number of columns in the table.
This implementation of SparseSVM functions solves the primal form of a linear kernel support vector machine, using gradient descent on the objective function. The implementation is based primarily on Pegasos: Primal Estimated Sub-Gradient Solver for SVM (by S. Shalev-Shwartz, Y. Singer, and N. Srebro; presented at ICML 2007).