Analysis of biologically plausible neuron models for regression with spiking neural networks

TL;DR

This study evaluates how different biologically plausible neuron models affect the accuracy and energy efficiency of Spiking Neural Networks in regression tasks, highlighting improvements over traditional models.

Contribution

It compares four neuron models within SNNs for regression, demonstrating that more biologically realistic models enhance accuracy and reduce energy consumption.

Findings

More realistic neuron models improve SNN regression accuracy.

Biologically plausible models reduce the number of spikes, saving energy.

Enhanced models show potential for neuromorphic hardware efficiency.

Abstract

This paper explores the impact of biologically plausible neuron models on the performance of Spiking Neural Networks (SNNs) for regression tasks. While SNNs are widely recognized for classification tasks, their application to Scientific Machine Learning and regression remains underexplored. We focus on the membrane component of SNNs, comparing four neuron models: Leaky Integrate-and-Fire, FitzHugh-Nagumo, Izhikevich, and Hodgkin-Huxley. We investigate their effect on SNN accuracy and efficiency for function regression tasks, by using Euler and Runge-Kutta 4th-order approximation schemes. We show how more biologically plausible neuron models improve the accuracy of SNNs while reducing the number of spikes in the system. The latter represents an energetic gain on actual neuromorphic chips since it directly reflects the amount of energy required for the computations.