Shot Noise Neuron Model

TL;DR

This paper introduces a shot noise-based leaky integrate-and-fire neuron model, providing a detailed performance comparison with traditional diffusion models, highlighting its biological realism and computational efficiency.

Contribution

It presents a novel shot noise neuron model and analyzes its performance relative to the diffusion approximation, demonstrating advantages in biological realism and computational efficiency.

Findings

The shot noise model captures spike timing behaviors more accurately.

It shows improved computational efficiency over traditional models.

The model exhibits phase transition and enhanced computational capacity in chaos.

Abstract

In this paper, we propose a shot noise-based leaky integrated and firing neuron model and provide a detailed analysis of the performance of this model compared to the traditional diffusion approximated model. In theoretical neuroscience, there are three general neuron models in the field: Compartmental neuron model is a conductance-based model, in which it views the biological neurons as a large circuit. The problem of this model comes from its structural complexity and the number of its free parameters; Leaky integrated and firing model is a more flexible model due to the special design called threshold-resetting, in which the voltage of the neuron is reset after reaching the threshold. This model is proposed as an alternative to the compartment model to provide a more biologically realistic model that can capture spike timing behaviors that are observed in experiments. In addition to that, it is more computationally efficient since it has much less free parameters; Firing rate-based neuron model uses the firing rate of the neuron as the coupling terms in the coupled neuronal network models. Models of this kind can be easily analyzed on the network level. Theoretical analysis for this model proves the existence of phase transition and shows that the computational capacity is boosted in the chaotic regime.