Attractor-merging Crises and Intermittency in Reservoir Computing

Abstract

Reservoir computing can embed attractors into random neural networks (RNNs), generating a ``mirror'' of a target attractor because of its inherent symmetrical constraints. In these RNNs, we report that an attractor-merging crisis accompanied by intermittency emerges simply by adjusting the global parameter. We further reveal its underlying mechanism through a detailed analysis of the phase-space structure and demonstrate that this bifurcation scenario is intrinsic to a general class of RNNs, independent of training data.

20 pages, 15 figures

References in corpus (12)

• Complexity without chaos: Plasticity within random recurrent networks generates robust timing and motor control
• Physical reservoir computing -- An introductory perspective
• Attractor Reconstruction by Machine Learning
• Embedding and Approximation Theorems for Echo State Networks
• Intermittent generalized synchronization in unidirectionally coupled chaotic oscillators
• Designing spontaneous behavioral switching via chaotic itinerancy
• Multifunctionality in a Reservoir Computer
• Optimal short-term memory before the edge of chaos in driven random recurrent networks
• Embedding bifurcations into pneumatic artificial muscle
• Finite-time Lyapunov exponents of deep neural networks
• Chaotic itinerancy and power-law residence time distribution in stochastic dynamical system
• Transient Chaos in BERT