Improved Long-Term Prediction of Chaos Using Reservoir Computing Based on Stochastic Spin-Orbit Torque Devices

TL;DR

This paper introduces a reservoir computing approach using stochastic spin-orbit torque devices to improve long-term predictions of chaotic systems like Mackey-Glass and Lorenz, leveraging their nonlinear and probabilistic properties.

Contribution

The study presents a novel RC system with SOT devices that enhances long-term chaotic prediction accuracy through stochastic nonlinear resistance changes.

Findings

Improved long-term prediction accuracy for chaotic systems.

Effective use of SOT devices' nonlinear and probabilistic properties.

Robustness of the RC network in chaotic system prediction.

Abstract

Predicting chaotic systems is crucial for understanding complex behaviors, yet challenging due to their sensitivity to initial conditions and inherent unpredictability. Probabilistic Reservoir Computing (RC) is well-suited for long-term chaotic predictions by handling complex dynamic systems. Spin-Orbit Torque (SOT) devices in spintronics, with their nonlinear and probabilistic operations, can enhance performance in these tasks. This study proposes an RC system utilizing SOT devices for predicting chaotic dynamics. By simulating the reservoir in an RC network with SOT devices that achieve nonlinear resistance changes with random distribution, we enhance the robustness for the predictive capability of the model. The RC network predicted the behaviors of the Mackey-Glass and Lorenz chaotic systems, demonstrating that stochastic SOT devices significantly improve long-term prediction accuracy.