Modelling with Relevance Vector Machines
Jan 2021
Jake Watson

The support vector machine (SVM) is a well-known technique for machine learning problems,
providing good generalisation with a sparse representation. It also has the disadvantages of
being unable to provide probabilistic outputs, and requiring its kernel functions to fulfil
stringent conditions. One approach to deal with these issues is the relevance vector machine
(RVM), a kernelized Bayesian general linear model, which provides similar generalisation
performance and increased sparsity, and includes probabilistic outputs by design.

This project is an R implementation of the RVM as stated by M.E. Tipping, [2] M. E. Tipping, “The Relevance Vector Machine,” in NIPS, Cambridge, 1999.