Performance of reservoir computing in a random network of single-walled carbon nanotubes complexed with polyoxometalate

Abstract

Molecular neuromorphic devices are composed of a random and extremely dense network of single-walled carbon nanotubes (SWNTs) complexed with polyoxometalate (POM). Such devices are expected to have the rudimentary ability of reservoir computing (RC), which utilizes signal response dynamics and a certain degree of network complexity. In this study, we performed RC using multiple signals collected from a SWNT/POM random network. The signals showed a nonlinear response with wide diversity originating from the network complexity. The performance of RC was evaluated for various tasks such as waveform reconstruction, a nonlinear autoregressive model, and memory capacity. The obtained results indicated its high capability as a nonlinear dynamical system, capable of information processing incorporated into edge computing in future technologies.

Edge-of-chaos learning achieved by ion-electron–coupled dynamics in an ion-gating reservoir

“…3, as a more challenging time series data analysis task. This is known as a NARMA2 task and is commonly used as a typical RC benchmark task ( 7 , 34 , 36 , 37 ). ynormaltfalse(k+1false)=0.4ynormaltfalse(kfalse)+0.4ynormaltfalse(kfalse)ynormaltfalse(k−1false)+0.6u3false(kfalse)+0.1where u ( k ) = [0,0.5] is a random input.…”

“…3 , as a more challenging time series data analysis task. This is known as a NARMA2 task and is commonly used as a typical RC benchmark task ( 7 , 34 , 36 , 37 ). where u ( k ) = [0,0.5] is a random input.…”

“…Furthermore, a strong dependence of the synaptic response on channel length, which is a feature of ion-electron-coupled dynamics, is used to realize high dimensionality in a single IGRT. The IGRT exhibited small errors in some RC tasks, including 0.020 of normalized mean squared error (NMSE) in a nonlinear autoregressive moving average (NARMA) task, which is a typical benchmark for RC (7,8,13,14,(34)(35)(36)(37), and achieved 88.8% accuracy in a handwritten-digit recognition task. The underlying mechanism of the characteristic synaptic response was investigated on the basis of multiphysics simulation, and it was found that complexed charge density patterns form and change from moment to moment in an extremely thin EDL region (<2 nm) during storage and processing of input signals.…”

In-materio computing in random networks of carbon nanotubes complexed with chemically dynamic molecules: a review

“…As an extension work of reference [29], reservoir performance of SWNT/POM was experimentally evaluated [68]. SWNT/POM with a polybutylmethacrylate (PBMA) composite was used to experimentally verify the RC task using NARMA-10.…”

Editorial: Focus on disordered, self-assembled neuromorphic systems

“…NARMA-10, waveform generation, memory capacity, Boolean logic gates) have been demonstrated using physical RC based on CNT network devices, and RC-based object classification has also been demonstrated in a robotic application. In a separate research article [2], Akai-Kasaya et al demonstrate how physical RC is implemented using the diverse nonlinear response signals from a dense SWCNT-POM network to perform waveform reconstruction and nonlinear autoregression, as well as the memory capacity task.…”