NOISE IN NEURAL NETWORKS — IN TERMS OF RELATIONS

Abstract

The issue of noise in biological systems is primarily a question of relations: between order and disorder, between stability and flexibility, and between processes at different temporal and spatial scales. In this paper, we use computational models of cortical structures to investigate relations between structure, dynamics, and function of such systems. In particular, we investigate the nature and role of noise at different organizational levels of the nervous system, emphasizing the neural network level. We show that microscopic noise can induce global network oscillations and pseudo-chaos, which make neural information processing more efficient. We find optimal noise levels for which the convergence to stored memory attractor states reaches a maximum, akin to stochastic resonance. We finally discuss the relation between neural and mental processes, and how computational models may relate to real neural systems.