Controlling Integration and Segregation in Echo State Networks via Noradrenaline and Acetylcholine Neuromodulation

Abstract

Biological brain networks flexibly reconfigure functional connectivity between integration and segregation through neuromodulatory systems--noradrenaline (NA) and acetylcholine (ACh)--without altering structural connectivity. Inspired by this mechanism, we propose a modular echo state network (ESN) with context-dependent NA and ACh gain modulation, where NA promotes inter-module integration via response gain and ACh promotes intra-module segregation via multiplicative gain. We evaluate the model on two context-dependent tasks: a segregation/integration task and a context-dependent decision task. Across both tasks, the modulated model consistently outperformed the baseline, with task-appropriate modulation profiles emerging naturally from optimization and functional connectivity analysis confirming context-appropriate dynamic reorganization. These results demonstrate that neuromodulatory gain control enables adaptive, context-sensitive computation in structurally fixed reservoir networks.