Gibbs distribution analysis of temporal correlations structure in retina ganglion cells

TL;DR

This paper introduces a method to estimate Gibbs distributions with spatio-temporal constraints for analyzing spike train data, providing a more accurate model of retinal ganglion cell activity than previous pairwise or Markov models.

Contribution

The authors develop a novel approach for modeling spatio-temporal correlations in neural spike trains, emphasizing higher order interactions and improving over existing models.

Findings

More accurate than pairwise synchronization models

Outperforms 1-time step Markov models

Highlights importance of higher order interactions

Abstract

We present a method to estimate Gibbs distributions with \textit{spatio-temporal} constraints on spike trains statistics. We apply this method to spike trains recorded from ganglion cells of the salamander retina, in response to natural movies. Our analysis, restricted to a few neurons, performs more accurately than pairwise synchronization models (Ising) or the 1-time step Markov models (\cite{marre-boustani-etal:09}) to describe the statistics of spatio-temporal spike patterns and emphasizes the role of higher order spatio-temporal interactions.