Structure and dynamics of random recurrent neural networks

Hugues Berry (1) , Mathias Quoy (2)

1 ALCHEMY - Architectures, Languages and Compilers to Harness the End of Moore Years
2 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes

In contradiction with Hopfield-like networks, random recurrent neural networks (RRNN), where the couplings are random, exhibit complex dynamics (limit cycles, chaos). It is possible to store information in these netwoks through hebbian learning. Eventually, learning ``destroys'' the dynamics and leads to a fixed point attractor. We investigate here the structural change in the networks through learning, and show a ``small-world'' effect.

Domaines: Réseau de neurones [cs.NE]; Biophysique [physics.bio-ph]; Systèmes désordonnés et réseaux de neurones [cond-mat.dis-nn]

Hugues Berry, Mathias Quoy. Structure and dynamics of random recurrent neural networks. Adaptive Behavior, 2006, 14, pp.129-137. ⟨inria-00001065⟩