On the information in spike timing: neural codes derived from polychronous groups

TL;DR

This paper investigates how spike timing encodes information in neural systems using a simple recurrent model that translates polychronous groups into a neural code, demonstrating properties similar to optimal codes and scalability with reservoir size.

Contribution

It introduces a minimalistic reservoir model that encodes spatiotemporal patterns into a neural code based on polychronous groups, analyzed with information theory.

Findings

Code distance properties resemble optimal random codes

Code achieves benchmarks for linear classification

Capacity scales exponentially with reservoir size

Abstract

There is growing evidence regarding the importance of spike timing in neural information processing, with even a small number of spikes carrying information, but computational models lag significantly behind those for rate coding. Experimental evidence on neuronal behavior is consistent with the dynamical and state dependent behavior provided by recurrent connections. This motivates the minimalistic abstraction investigated in this paper, aimed at providing insight into information encoding in spike timing via recurrent connections. We employ information-theoretic techniques for a simple reservoir model which encodes input spatiotemporal patterns into a sparse neural code, translating the polychronous groups introduced by Izhikevich into codewords on which we can perform standard vector operations. We show that the distance properties of the code are similar to those for (optimal) random codes. In particular, the code meets benchmarks associated with both linear classification and capacity, with the latter scaling exponentially with reservoir size.