Structural determinants of soft memory in recurrent biological networks

TL;DR

This paper investigates how the structural features of biological recurrent networks, especially gene regulatory networks in bacteria, influence their ability to store and process information dynamically.

Contribution

It introduces a novel analysis of biological network topology, linking structural properties to information storage and processing capabilities.

Findings

Certain topological features enhance information storage.

Global and local structures influence network dynamics.

Biological networks are optimized for information processing.

Abstract

Recurrent neural networks are frequently studied in terms of their information-processing capabilities. The structural properties of these networks are seldom considered, beyond those emerging from the connectivity tuning necessary for network training. However, real biological networks have non-contingent architectures that have been shaped by evolution over eons, constrained partly by information-processing criteria, but more generally by fitness maximization requirements. Here we examine the topological properties of existing biological networks, focusing in particular on gene regulatory networks in bacteria. We identify structural features, both local and global, that dictate the ability of recurrent networks to store information on the fly and process complex time-dependent inputs.