Multifunctional asynchronous reservoir computing system with polarization-controlled dual optoelectronic feedback loops
Zhiyi Xu, Bo Yang, Shuna Yang, Hao Chi · Optical Engineering · 2025
We propose a multifunctional asynchronous reservoir computing system based on polarization-controlled dual optoelectronic feedback loops with a dual-polarization binary phase-shift keying modulator. By incorporating a polarization controller for mode switching, the system supports both parallel dual-task mode and enhanced single-task mode. In the enhanced single-task mode, the system outperforms traditional single-loop systems in prediction and classification accuracy through asymmetric coupling, asynchronous processing, and adjustable coupling strength. In parallel dual-task mode, different tasks are trained in separate feedback loops, independently optimizing nonlinear functions to enhance the performance of each task. Numerical simulations verify the feasibility of the proposed scheme. In enhanced single-task mode, the normalized mean squared error (NMSE) for the nonlinear autoregressive moving average (NARMA10) task is reduced by 0.0958, and the symbol error rate (SER) for nonlinear channel equalization is significantly improved compared with traditional single-loop systems. In addition, when processing NARMA10 and IPIX radar signal prediction tasks, the NMSE in parallel dual-task mode is reduced by 0.046 and 0.0135, respectively, compared with single-loop processing.