Predicting with Relevance Vector Machines.

I am trying out this Matlab toolbox for Relevance Vector Machines by Tipping: http://www.miketipping.com/sparsebayes.htm

This has an implementation of Relevance Vector Machines, and generates pretty nice figures. But what it doesn't talk about is classification accuracy of the model it has returned after training and of course the exact equation one must use for it.

On line 242, it trains the classifier in the script SparseBayesDemo:

[PARAMETER, HYPERPARAMETER, DIAGNOSTIC] = ... SparseBayes(likelihood_, BASIS, Outputs, OPTIONS, SETTINGS)

Has anyone any experience here implementing RVM for classification tasks and can assist? I will be grateful for a response.