Parallel reservoir computing system based on a WRC-FPLD subject to optical feedback with electrical information injection

• Published in: Optics communications Vol. 594; p. 132345
• Main Authors: Zhang, Heman, Wang, Qiupin, Gao, Xulin, Ou, Pu, Lei, Zhiqiang, Tong, Xin, Lu, Dan, Wu, Zhengmao, Xia, Guangqiong
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.11.2025
• Subjects: Electrical information injection; Optical feedback; Parallel reservoir computing; Weak resonant-cavity Fabry-Perot laser diode (WRC-FPLD); Parallel reservoir computing; Optical feedback; Electrical information injection; Weak resonant-cavity Fabry-Perot laser diode (WRC-FPLD)
• ISSN: 0030-4018
• DOI: 10.1016/j.optcom.2025.132345

In this work, we propose and experimentally demonstrate a parallel reservoir computing (RC) system based on a weak resonant-cavity Fabry-Perot laser diode (WRC-FPLD) under optical feedback with electrical information injection. This system consists of an input layer, a reservoir layer, and an output layer. In the input layer, the information that needs to be processed is loaded on the bias current of WRC-FPLD through current modulation. In the reservoir layer, WRC-FPLD is served as a nonlinear node, and the optical feedback loop is taken as a virtual node net with a uniform time interval. In the output layer, by sampling virtual node states from either single-mode or multi-mode for training and testing, the output of the RC system can be obtained. Via Santa-Fe chaotic time series prediction benchmark task, the performance of this RC system is evaluated. The results show that, comparing extracting single-mode with extracting multi-mode as the states of virtual node, the former can achieve a minimum normalized mean square error (NMSE) of 0.0502 with the data processing rate (DPR) of 12.50 MSa/s, and the latter can further increase DPR to 39.00 MSa/s with a NMSE below 0.1. This approach presents a promising direction for developing high-speed RC systems with improved error tolerance.