One dimensional Fokker-Planck reduced dynamics of decision making models in Computational Neuroscience

TL;DR

This paper derives a simplified one-dimensional Fokker-Planck equation from a complex two-population neural model, enabling efficient analysis of equilibrium states and dynamics in computational neuroscience.

Contribution

It introduces a novel reduction method leveraging slow-fast dynamics to simplify the Fokker-Planck model of neural populations.

Findings

The 1D reduced model accurately captures equilibrium states.

The reduced model provides insights into escape times and parameter dependencies.

Validation shows the reduced model aligns well with the original 2D system.

Abstract

We study a Fokker-Planck equation modelling the firing rates of two interacting populations of neurons. This model arises in computational neuroscience when considering, for example, bistable visual perception problems and is based on a stochastic Wilson-Cowan system of differential equations. In a previous work, the slow-fast behavior of the solution of the Fokker-Planck equation has been highlighted. Our aim is to demonstrate that the complexity of the model can be drastically reduced using this slow-fast structure. In fact, we can derive a one-dimensional Fokker-Planck equation that describes the evolution of the solution along the so-called slow manifold. This permits to have a direct efficient determination of the equilibrium state and its effective potential, and thus to investigate its dependencies with respect to various parameters of the model. It also allows to obtain information about the time escaping behavior. The results obtained for the reduced 1D equation are validated with those of the original 2D equation both for equilibrium and transient behavior.