Greenberg-Hastings dynamics on a small-world network: the collective extinct-active transition

TL;DR

This study investigates how a reaction-diffusion model on small-world networks transitions between extinct and active states, revealing parameter relations that control this transition and implications for systems operating at criticality.

Contribution

The paper provides a numerical analysis of the extinct-active transition on small-world networks, deriving explicit relations between network parameters and the transition, and discusses implications for information processing.

Findings

Transition controlled by network parameters $K$ and $p$

Transition can be induced by changing $r$ or $K$

Model operates at criticality for optimal information processing

Abstract

We present a numerical study of a reaction-diffusion model on a small-world network. We characterize the model's average activity $F_T$ after $T$ time steps and the transition from a collective (global) extinct state to an active state in parameter space. We provide an explicit relation between the parameters of our model at the frontier between these states. A collective active state can be associated to a global epidemic spread, or to a persistent neuronal activity. We found that $F_T$ does not depends on disorder in the network if the transmission rate $r$ or the average coordination number $K$ are large enough. The collective extinct-active transition can be induced by changing two parameters associated to the network: $K$ and the disorder parameter $p$ (which controls the variance of $K$). We can also induce the transition by changing $r$, which controls the threshold size in the dynamics. In order to operate at the transition the parameters of the model must satisfy the relation $rK=a_p$, where $a_p$ as a function of $p/(1-p)$ is a stretched exponential function. Our results are relevant for systems that operate {\it at} the transition in order to increase its dynamic range and/or to operate under optimal information-processing conditions. We discuss how glassy behaviour appears within our model.