The memristor-based chaotic system (MCS) exhibits extreme sensitivity to initial conditions, rendering accurate prediction of its behavior highly challenging. This study integrates an enhanced reservoir computing (RC) approach with the Ivy algorithm (IVYA) so as to propose a novel prediction framework for MCS, termed IVYA-based one-dimensional RC (IVYA-ODRC). Unlike traditional approaches requiring precise mathematical models, IVYA-ODRC operates without prior knowledge of the system. Additionally, while conventional RC methods typically perform only single-step or autonomous prediction, IVYA-ODRC can accurately predict long-term dynamics using only a single state variable. It offers an innovative approach for prediction tasks under low-dimensional observation scenarios. Remarkably, the IVYA-ODRC method effectively mitigates error accumulation, maintaining robust prediction performance over extended periods. Given a sufficiently long sequence of observations, the method maintains stable performance over extended periods. Moreover, experimental results demonstrate the robustness of IVYA-ODRC under high levels of stochastic noise, with strong performance maintained when noise is introduced in the prediction phase as well as in both training and prediction phases.
Wang et al. (Fri,) studied this question.