Experimental study on parallel and analog optical reservoir computing with delayed feedback system for physical implementation

Abstract

Optical reservoir computing (RC) with delayed feedback is expected to achieve high-speed data processing. However, in the parallel RC framework, the digital pre-process limits the actual processing speed. An analog-based, simple pre-processing method was developed and implemented in an optical RC architecture to overcome the bottleneck, and the performance in calculating two types of parallel task was evaluated. One was two independent benchmark tasks; the other was an integrative multi-input and multi-output odor identification task. We successfully demonstrated that these two parallel tasks can be physically processed with the network parameters optimized to maximize the RC performance. These results strongly suggest the potential of the optical reservoir system for future high-speed multimodal data processing applications.