A Markov random field approach to neural encoding and decoding

Abstract

We introduce a new approach to neural encoding and decoding which makes use of sparse regression and Markov random fields. We show that interesting response functions were estimated from neuroimaging data acquired while a subject was watching checkerboard patterns and geometrical figures. Furthermore, we demonstrate that reconstructions of the original stimuli can be generated by loopy belief propagation in a Markov random field. © 2011 Springer-Verlag.