[10a-S101-1~9] AI Accelerator: Next stage of artificial intelligence devices

[10a-S101-6] Reservoir computing using ferroelectric-based devices

〇Shinichi Takagi¹, Kasidit Toprasertpong¹, Eishin Nako¹, Zeyu Wang¹, Mitsuru Takenaka¹, Ryosho Nakane¹ (1.U. Tokyo)

Keywords:reservoir computing, Ferroelectrics

Physical reservoir computing, where fast learning computation can be executed with high efficienty, is attracting much attention for applications of edge AI computing. In this presentation, we will introduce the advantages and concrete methods of reservoir computing using HfO2-based ferroelectric gate insulator FETs (FeFETs) and MFM capacitors with temporary memory and nonlinearity as a physical reservoir, which we have recently proposed. We will show the experimental results to verify that AI tasks can be realized by reservoir computing utilizing the ferroelectric properties.