Reservoir computer using a photorefractive nonlinear medium

Reservoir computing (RC) is a machine learning (ML) model that has gained popularity in the recent decade due to its simplicity, versatility in solving temporal problems, and ease of implementation across various physical systems. This work explores the implementation of a photorefractive crystal as the reservoir in a reservoir computer. While previous studies have demonstrated the feasibility of using photorefractive materials for RC and other ML applications, a systematic investigation into the impact of the optical nonlinearity on the systems performance has yet to be conducted. To address this question, we develop a reservoir computer using a biased photorefractive crystal. By quantifying the system’s optical nonlinearity and evaluating the performance of the reservoir computer, we reveal a correlation between these factors, providing new insights into optimizing photorefractive systems for machine learning tasks.