The support vector machine algorithm has a single hyperparameter C that regularizes the learned model. Recently, Hastie ## A fast regularization pathway for the support vector machine

## Oliver Serang and William Stafford Noble

Abstractet al.(2004) described an algorithm for finding the SVM solution for all possible values of this regularization parameter. We present an efficient C++ implementation of this algorithm called svmvia. We compare svmvia running time to that of libsvm on three simulated and two real data sets. Depending upon the data set, for small values of C svmvia can take approximately 10-1000 times as long as libsvm. However, for large values of C, it is often faster to find the entire regularization path than to train a single model. svmvia is available at http://noble.gs.washington.edu/proj/svmvia.

Download svmvia here.

Download data files here.

A guide to using svmvia is here.

svmvia was created by Oliver Serang under the guidance of Prof. William Stafford Noble in the Department of Genome Sciences at the University of Washington.

svmvia is released under a GPL 2.0 software license.