Presentation Information
[10p-PA3-21]Reservoir computing using magnetization dynamics of random magnets
〇Akane Onodera1, Shuto Kamakura1, Tomi Ohtsuki2, Jun-ichiro Ohe1 (1.Toho University, 2.Sophia University)
Keywords:
Spintronics,Machine Learning
Physical reservoir computing (RC) utilizing spintronics remains challenging due to the necessity of advanced microfabrication techniques. In this study, we propose an easily fabricable magnetic material containing magnetic impurities, and numerically verify its RC capability by exploiting magnetization dynamics dominated by dipole-dipole interactions. Our results demonstrate that this system can perform chaotic time-series prediction and classify localized/delocalized electronic states. We also reveal how the strength of long-range interactions significantly enhances RC performance.