Scalable photonic reservoir computing based on pulse propagation in parallel passive dispersive links

Abstract

We propose and demonstrate a scalable photonic reservoir computing system based on pulse propagation in parallel passive dispersive links. The approach features a simple structure and is composed of passive components, such as dispersive elements and ring resonators. The system utilizes a pulsed laser and employs a multi-channel architecture with dispersive elements for pulse broadening and ring resonators for delay stacking. It is verified for its capability to handle both digital tasks (2-bit Boolean logic operations, header recognition) and analog tasks (Mackey-Glass time-series prediction, Lorenz 63 time-series prediction). A major advantage of the approach lies in its scalability. It has been confirmed that more complex tasks can be processed by using a system with more channels, along with dispersive elements and ring resonators with larger dispersion amounts and delays, respectively.