Multiple Simultaneous Pseudo Image Classification with Random Fields and a Deep Belief Network for Disease Indication

We show how to use random field theory in a supervised, energy-based model for multiple pseudo image classification of 2D integer matrices. In the model, each row of a 2D integer matrix is a pseudo image where a local receptive field focuses on multiple portions of individual rows for simultaneous learning. The model is used for a classification task consisting of presence of patient biomarkers indicative of a particular disease.