Hardware optimization for photonic time-delay reservoir computer dynamics

Abstract Reservoir computing (RC) is one kind of neuromorphic mainly applied to process sequential data such as time-dependent signals. In this paper, the bifurcation diagram a photonic time-delay RC system thoroughly studied, and method dynamics guided hardware hyperparameter optimization presented. The time-evolution equation expressed by parameters established while intrinsic quantitively studied. Bifurcation based offers simple yet effective approach in setting that aims reduce complexity time adjustment. Three benchmark tasks, nonlinear channel equalization (NCE), auto regressive moving average with 10th order lag (NARMA10) Santa Fe laser time-series prediction tasks are implemented on delay-line using optimization. experimental results these achieved overall good agreement simulated modeling results.