Presentation Information
[9a-B21-2]Physical reservoir computing based on redox-active gel/PEDOT:PSS laminated devices
〇Toshiyuki Tanaka1, Kato Hiroyuki1, Akai Megumi1, Watanabe Seiya1 (1.The Univ. of Osaka)
Keywords:
physical reservoir computing
Physical reservoir computing (PRC) uses a material's nonlinear response and short-term memory as computational resources, offering a low-power route to edge AI. We investigate the PRC capability of an electrochemical device of a redox-active gel laminated with PEDOT:PSS. Doping/dedoping of PEDOT:PSS and the redox reaction of ferrocene in the gel produce nonlinear, asymmetric current responses. Under random step-voltage inputs, we extracted virtual nodes from the current responses and built a multi-node reservoir by multiplexing responses from different potential ranges and from PEDOT-coated and bare electrodes. Multiplexing substantially increased the memory capacity, which is dominated by nonlinear components, indicating that the nonlinear electrochemical response drives the computing performance. We will also report simultaneous multi-point measurements.