Collective oscillations in disordered neural networks

TL;DR

This paper studies how disorder in neural networks can induce weak chaos and affect collective oscillations, revealing that in large networks the dynamics resemble a homogeneous system with scaled coupling.

Contribution

It demonstrates that disorder induces a weak form of chaos in neural networks and shows how the Lyapunov spectrum scales with network size, connecting to known states like the splay state.

Findings

Disorder induces a weak form of chaos in neural networks.

Lyapunov exponent scales to zero as network size increases.

In the thermodynamic limit, dynamics resemble a homogeneous system.

Abstract

We investigate the onset of collective oscillations in a network of pulse-coupled leaky-integrate-and-fire neurons in the presence of quenched and annealed disorder. We find that the disorder induces a weak form of chaos that is analogous to that arising in the Kuramoto model for a finite number N of oscillators [O.V. Popovych at al., Phys. Rev. E 71} 065201(R) (2005)]. In fact, the maximum Lyapunov exponent turns out to scale to zero for N going to infinite, with an exponent that is different for the two types of disorder. In the thermodynamic limit, the random-network dynamics reduces to that of a fully homogenous system with a suitably scaled coupling strength. Moreover, we show that the Lyapunov spectrum of the periodically collective state scales to zero as 1/N^2, analogously to the scaling found for the `splay state'.