[23a-M206-1~10] FS.1 Focused Session "AI Electronics"

△ [23a-M206-8] Effect of sampling step in chaotic time series regenerated by reservoir computing

〇(M1)Kohei Tsuchiyama¹, Andre Roehm¹, Takatomo Mihana¹, Ryoichi Horisaki¹, Makoto Naruse¹ (1.Univ. of Tokyo)

Keywords:Reservoir computing, Chaos, Sampling

Reservoir computing (RC) has been intensively studied as a highly efficient machine learning approach utilizing physical processes. Recently, autonomous reproduction of chaotic time series has been demonstrated by RC. Signals, such as time series, are always sampled.In this study, we examine the impact of sampling intervals on the ability to reproduce chaotic time series. We demonstrate that the reproduction ability suddenly drops beyond a certain sampling interval based on numerical simulations.