Modular organization enhances the robustness of attractor network dynamics

TL;DR

This paper demonstrates that modular organization in neural networks enhances the robustness and stability of attractor states, leading to faster convergence and larger basins of attraction, which may explain its prevalence in natural systems.

Contribution

It shows that modular structure increases the robustness of attractor dynamics in neural networks, revealing a potential evolutionary advantage of modularity in complex systems.

Findings

Maximum volume of attractor basins at optimal modularity

Decreased convergence time with increased modularity

Modularity contributes to global stability of attractors

Abstract

Modular organization characterizes many complex networks occurring in nature, including the brain. In this paper we show that modular structure may be responsible for increasing the robustness of certain dynamical states of such systems. In a neural network model with threshold-activated binary elements, we observe that the basins of attractors, corresponding to patterns that have been embedded using a learning rule, occupy maximum volume in phase space at an optimal modularity. Simultaneously, the convergence time to these attractors decreases as a result of cooperative dynamics between the modules. The role of modularity in increasing global stability of certain desirable attractors of a system may provide a clue to its evolution and ubiquity in natural systems.