Optical reservoir computing with a semiconductor laser based on a multi-feedback fiber-optic ring structure: system design and performance analysis

Optical reservoir computing (ORC) offers efficient training and simple hardware implementation, making it suitable for time-series forecasting and related tasks. Single-feedback ORC structures have limited dynamic capacity, while multi-feedback designs improve performance but increase structural complexity. This work proposes a fiber-optic ring (FOR)-based ORC system that leverages the intrinsic delay characteristics of the FOR to form multiple feedback loops within a compact and easily implementable architecture. Simulation results demonstrate that the proposed system significantly surpasses dual-feedback ORC and single-feedback ORC in memory capacity (MC) and time-series forecasting tasks. Specifically, the MC of the proposed ORC is comparable to that of deep ORC with 2–3 cascaded layers, which indicates that the proposed ORC has strong potential for achieving multi-feedback RC while maintaining as a simple structure.