Neural Population Coding is Optimized by Discrete Tuning Curves

TL;DR

This paper demonstrates that optimal neural coding with Poisson neurons involves discrete tuning curves with quantized input levels, and reveals phase transitions in the number of quantization levels as coding parameters change.

Contribution

It shows that the optimal tuning curves are discrete and that the number of quantization levels undergoes phase transitions, providing insights into neural population coding strategies.

Findings

Optimal tuning curves are discrete with quantized input levels.

Number of quantization levels exhibits phase transitions with coding window length.

Subpopulation structure in neural populations aligns with optimal coding principles.

Abstract

The sigmoidal tuning curve that maximizes the mutual information for a Poisson neuron, or population of Poisson neurons, is obtained. The optimal tuning curve is found to have a discrete structure that results in a quantization of the input signal. The number of quantization levels undergoes a hierarchy of phase transitions as the length of the coding window is varied. We postulate, using the mammalian auditory system as an example, that the presence of a subpopulation structure within a neural population is consistent with an optimal neural code.