A reinforced reservoir computer aided by an external asymmetric dual-path-filtering cavity laser

Abstract

Chaos, characterized by irregular, stochastic-like and occurring in deterministic systems, is widely used in meteorology, life sciences, and physics. Precise chaos predictions are crucial for early warning of extreme weather and disease prevention. We propose a photonic time-delay reservoir computing (TDRC) system with asymmetric dual-path filtering optical feedback under optical injection for short-term prediction of chaotic time series. To thoroughly evaluate the performance in short-term prediction provided by such TDRC, we assess two different chaotic time series, i.e., the Santa-Fe and Mackey-Glass chaotic time series, as well as the memory capacity. Numerical results indicate that the proposed TDRC outperforms the system with conventional dual-path optical feedback in short-term prediction performance. This is attributed to the enhanced memory capacity originating from the asymmetric dual-path filtering optical feedback. Additionally, we reveal the effects of the injection strength, feedback strength, filter bandwidth and the number of virtual nodes on the system performance. Our work provides a novel path for accurate short-term prediction of complex chaotic systems using photonic TDRC.