[9a-Z24-1~10] 【CS.7】 Code-sharing Session of 7.4 & 9.5

▼ [9a-Z24-10] Physical reservoir system with single-walled carbon nanotube/ asymmetric porphyrin-sandwiched polyoxometalate random network

〇(D)Deep Banerjee¹, Takumi Kotooka¹, Yoshito Yamazaki², Takuji Ogawa², Hirofumi Tanaka¹ (1.KYUTECH, 2.Osaka Univ)

Keywords:Reservoir computing, Single walled carbon nanotube, Porphyrin polyoxometalate

Reservoir computing (RC) has become a core deep neural network algortihm for learning and training big complex data in an energy-efficient way. Among various other physical reservoir systems already reported we here exploit the dynamics of single-waaled carbon nanitube/porphyrin polyoxometalate (SWNT/Por-POM) by conducting electrical measurements. Properties like non-linearity, inverse power law scaling, fading memory, high dimensionality and phase delayed non-linear outputs (echo-state property) from a time varying sine wave input satisfies the RC charecteristics. The benchmark RC task of waveform generation and non-linear autoregressive moving average (NARMA) time prediction series were also successfully achieved. Based on these findings we conclude SWNT/Por-POM can solve cognitive tasks in real time and can be used in the future for speech recognition as well.