Asymptotic behavior of a network of neurons with random linear interactions

TL;DR

This paper analyzes the long-term behavior of a neural network with random linear interactions, showing convergence to a Gaussian process described by a non-Markovian process, using the method of moments.

Contribution

It introduces a new asymptotic analysis of asymmetric neural networks with random couplings, characterizing the limit as a Gaussian process with explicit covariance.

Findings

Limit system described by Gaussian process with Bessel function covariance

Almost sure convergence in law with respect to random weights

Method of moments used to establish Central Limit Theorem

Abstract

We study the asymptotic behavior for asymmetric neuronal dynamics in a network of linear Hopfield neurons. The interaction between the neurons is modeled by random couplings which are centered i.i.d. random variables with finite moments of all orders. We prove that if the initial condition of the network is a set of i.i.d. random variables and independent of the synaptic weights, each component of the limit system is described as the sum of the corresponding coordinate of the initial condition with a centered Gaussian process whose covariance function can be described in terms of a modified Bessel function. This process is not Markovian. The convergence is in law almost surely with respect to the random weights. Our method is essentially based on the method of moments to obtain a Central Limit Theorem.