Constructive Role of Noise for High-Quality Replication of Chaotic Attractor Dynamics Using a Hardware-Based Reservoir Computer
Irene Estébanez, Ingo Fischer, Miguel C. Soriano · Physical Review Applied · 2019
The identification of unconventional physical systems for neuroinspired information processing comes with much-desired capabilities for powerful data analysis. Moreover, it offers the possibility to replicate complex dynamical behavior under closed-loop operation. Robust, reliable autonomous replication of the structure of complex dynamics has remained elusive, though. The authors experimentally demonstrate the advantage of pretraining a system with additive input noise, achieving robust operation over an extended parameter range, and introduce quantifiers for the quality of replication.