[19p-Z34-1~17] FS.1 フォーカストセッション「AIエレクトロニクス」

▼ [19p-Z34-7] Reservoir computing enhancement in three-dimensional porous CNT-POM network

〇(P)Saman Azhari^1,2、Deep Banerjee¹、Takumi Kotooka¹、Yuki Usami^1,2、Hirofumi Tanaka^1,2 (1.Graduate School of Life Science and Systems Engineering (Human Intelligence Systems), Kyushu Institute of Technology (Kyutech)、2.Research Center for Neuromorphic AI Hardware, Kyushu Institute of Technology (Kyutech))

キーワード：Recurrent Neural Network, Reservoir Computing, Three-Dimensional Network

A three-dimensional porous template of melamine-cellulose fiber is utilized to devise a three-dimensional network of CNT-POM network for enhanced reservoir computing performance. Although all samples exhibit higher dimensionality and non-linearity, the sample with higher cellulose content and increased template's porosity exhibit superior reservoir performance due to recurrent network formation. Our results indicate the importance of device topology and network distribution on the performance of physical reservoir computing devices.