Edge-of-Chaos and Chaotic Dynamics in Resistor-Inductor-Diode-Based Reservoir Computing

Series Resistor-Inductor-Diode (RLD) circuits are known to exhibit rich nonlinear dynamical behaviour that opens up intriguing opportunities for design of novel computational systems. In this paper, we suggest and theoretically validate a technically simple RLD circuit that implements a reservoir computing (RC) architecture optimized to predict the future evolution of highly nonlinear and chaotic time series. We demonstrate that the exploitation of the proposed RLD circuit in a periodic operating regime with a particular tuning on an edge-of-chaos mode enables the RLD-based reservoir computer to achieve high prediction accuracy, quantified as a normalized mean square error (NMSE) of approximately $10^{-4}$ . We also evaluate the performance of RLD-RC tuned to operate in a chaotic regime under forced initial conditions, revealing the ability of the so-designed computer to accurately forecast complex time series and highlighting the potential of RLD circuits to serve as a backbone of efficient and versatile hardware RC systems.