Classification of medical data using Restricted Boltzmann Machines

Abstract

Restricted Boltzmann Machines are generative models commonly used for feature extraction and for training deep neural networks. In this thesis, their applicability for classification of medical data is researched. Three different approaches are evaluated using two small medical data sets. It is shown that the resulting classifiers are able to form sensible models of the data, having competitive performance when compared to other methods on these data sets.