The Japan Society of Applied Physics

11:15 〜 11:30

[B-6-02] Deep Physical Reservoir Computing Achieved by Electric Double Layer Ion-Gating Reservoir

Daiki Nishioka1,2, Takashi Tsuchiya1, Wataru Namiki1, Masataka Imura1, Yasuo Koide1, Tohru Higuchi2, Kazuya Terabe1 (1. National Institute for Materials Science (Japan), 2. Tokyo Univ. of Sci. (Japan))

https://doi.org/10.7567/SSDM.2023.B-6-02

In order to boost computation performance of physical reservoir computing (PRC), we physically implemented deep reservoir computing on an ion-gating reservoir; PRC nanodevice, which utilizes ion-electron coupled dynamics in an electric double layer at the Li+ solid electrolyte/Hdiamond interface. The deep PRC achieved a normalized mean squared error of 0.0141 on the second-order nonlinear autoregressive moving average (NARMA2) task, the best performance of any reported physical reservoir nanodevice, and also outperformed full simulation reservoir computing performance.