[Paper Review] A Generalized Rate Model for Neuronal Ensembles

Hideo Hasegawa|arXiv (Cornell University)|Mar 20, 2007

Neural dynamics and brain function6 references3 citations

TL;DR

This paper proposes a generalized rate model for neuronal ensembles using a finite-N Langevin framework with additive and multiplicative noise, enabling analysis of both stationary and dynamical properties via the Fokker-Planck equation and augmented moment method (AMM). The key contribution is demonstrating that population rate coding is more reliable than single-neuron coding due to $ρ \sim \gamma/N$, where $ρ$ is global fluctuation and $γ$ is local fluctuation, with AMM predictions matching direct simulations for pulse and sinusoidal inputs.

ABSTRACT

There has been a long-standing controversy whether information in neuronal networks is carried by the firing rate code or by the firing temporal code. The current status of the rivalry between the two codes is briefly reviewed with the recent studies such as the brain-machine interface (BMI). Then we have proposed a generalized rate model based on the {\it finite} $N$-unit Langevin model subjected to additive and/or multiplicative noises, in order to understand the firing property of a cluster containing $N$ neurons. The stationary property of the rate model has been studied with the use of the Fokker-Planck equation (FPE) method. Our rate model is shown to yield various kinds of stationary distributions such as the interspike-interval distribution expressed by non-Gaussians including gamma, inverse-Gaussian-like and log-normal-like distributions. The dynamical property of the generalized rate model has been studied with the use of the augmented moment method (AMM) which was developed by the author [H. Hasegawa, J. Phys. Soc. Jpn. 75 (2006) 033001]. From the macroscopic point of view in the AMM, the property of the $N$-unit neuron cluster is expressed in terms of {\it three} quantities; $μ$, the mean of spiking rates of $R=(1/N) \sum_i r_i$ where $r_i$ denotes the firing rate of a neuron $i$ in the cluster: $γ$, averaged fluctuations in local variables ($r_i$): $ρ$, fluctuations in global variable ($R$). We get equations of motions of the three quantities, which show $ρ\sim γ/N$ for weak couplings. This implies that the population rate code is generally more reliable than the single-neuron rate code. Our rate model is extended and applied to an ensemble containing multiple neuron clusters.

Motivation & Objective

• To resolve the longstanding debate on whether neural information is encoded by firing rate or temporal codes in neuronal networks.
• To develop a unified rate model that captures both stationary and dynamical properties of finite-size neuronal ensembles.
• To investigate the reliability of population rate coding versus single-neuron rate coding in the presence of noise and finite-size effects.
• To extend the model to multi-cluster systems, particularly a generalized Wilson-Cowan model with excitatory and inhibitory populations.

Proposed method

• Formulates a finite-N-unit Langevin model with additive and multiplicative noise to describe spiking dynamics in neuronal clusters.
• Applies the Fokker-Planck equation (FPE) to analyze stationary distributions of interspike intervals, yielding non-Gaussian forms such as gamma, inverse-Gaussian-like, and log-normal-like distributions.
• Employs the augmented moment method (AMM) to derive macroscopic equations of motion for three key variables: mean rate ($\mu$), local fluctuation ($\gamma$), and global fluctuation ($\rho$).
• Derives coupled differential equations for $\mu_m$, $\gamma_m$, and $\rho_{mn}$ by expanding around mean values and retaining second-order cumulants, enabling analysis of collective dynamics.
• Validates the AMM predictions against direct simulations (DS) for responses to pulse and sinusoidal inputs, showing strong agreement.
• Extends the model to multi-cluster ensembles, particularly a two-cluster system of excitatory and inhibitory neurons, using weighted coupling terms and external inputs.

Experimental results

Research questions

• RQ1 How do stationary interspike interval distributions emerge in finite neuronal ensembles under noise, and what forms do they take?
• RQ2 What is the relationship between local fluctuations ($\gamma$) and global population fluctuations ($\rho$) in neuronal ensembles, and how does it affect coding reliability?
• RQ3 How accurately can the augmented moment method (AMM) predict the dynamical response of neuronal clusters to external inputs such as pulses and sinusoids?
• RQ4 How does the generalized rate model perform when extended to multi-cluster systems, such as excitatory-inhibitory neuronal ensembles?

Key findings

• The generalized rate model produces stationary interspike interval distributions that include non-Gaussian forms such as gamma, inverse-Gaussian-like, and log-normal-like distributions, consistent with experimental observations.
• The AMM reveals that global fluctuation $\rho$ scales as $\gamma/N$ for weak couplings, indicating that population rate coding is inherently more reliable than single-neuron rate coding.
• Dynamical responses of the model to pulse and sinusoidal inputs, as predicted by the AMM, show excellent agreement with results from direct simulations (DS).
• The model successfully generalizes the Wilson-Cowan framework to multi-cluster ensembles, with explicit equations of motion derived for mean, variance, and covariance of global rates in two-cluster systems.
• The derived equations for $\mu_m$, $\gamma_m$, and $\rho_{mn}$ include contributions from nonlinearities in the rate function $F$, noise terms $\alpha_m^2$ and $\beta_m^2$, and coupling weights $w_{mn}$, enabling quantitative analysis of collective dynamics.