Search

While physical reservoir computing (PRC) is a promising way to achieve low power consumption neuromorphic computing, its computational performance is still insufficient at a practical level. One promising approach to improving PRC performance is deep reservoir computing (deep-RC), in which the component reservoirs are multi-layered. However, all of the deep-RC schemes reported so far have been effective only for simulation reservoirs and limited PRCs, and there have been no reports of nanodevice implementations. Here, as the first nanodevice implementation of Deep-RC, we report a demonstration of deep physical reservoir computing using an ion gating reservoir (IGR), which is a small and high-performance physical reservoir. While previously reported Deep-RC scheme did not improve the performance of IGR, our Deep-IGR achieved a normalized mean squared error of 0.0092 on a second-order nonlinear autoregressive moving average task, with is the best performance of any physical reservoir so far reported. More importantly, the device outperformed full simulation reservoir computing. The dramatic performance improvement of the IGR with our deep-RC architecture paves the way for high-performance, large-scale, physical neural network devices., Comment: 21 pages, 6 figures

Reservoir computing (RC) is a promising solution for achieving low power consumption neuromorphic computing, although the large volume of the physical reservoirs reported to date has been a serious drawback in their practical application. Here, we report the development of a few-molecule RC that employs the molecular vibration dynamics in the para-mercaptobenzoic acid (pMBA) detected by surface enhanced Raman scattering (SERS) with tungsten oxide nanorod/silver nanoparticles (WOx@Ag-NPs). The Raman signals of the pMBA molecules, adsorbed at the SERS active site of WOx@Ag-NPs, were reversibly perturbated by the application of voltage-induced local pH changes in the vicinity of the molecules, and then used to perform RC of pattern recognition and prediction tasks. In spite of the small number of molecules employed, our system achieved good performance, including 95.1% to 97.7% accuracy in various nonlinear waveform transformations and 94.3% accuracy in solving a second-order nonlinear dynamic equation task. Our work provides a new concept of molecular computing with practical computation capabilities., Comment: 22 pages, 4 figures

We have demonstrated physical reservoir computing with a redox-based ion-gating reservoir (redox-IGR) comprising LixWO3 thin film and lithium ion conducting glass ceramic (LICGC). The subject redox-IGR successfully solved a second-order nonlinear dynamic equation by utilizing voltage pulse driven ion-gating in a LixWO3 channel to enable reservoir computing. Under the normal conditions, in which only the drain current (ID) is used for the reservoir states, the lowest prediction error is 7.39x10-4. Performance was enhanced by the addition of IG to the reservoir states, resulting in a significant lowering of the prediction error to 5.06x10-4, which is noticeably lower than other types of physical reservoirs reported to date. A second-order nonlinear autoregressive moving average (NARMA2) task, a typical benchmark of reservoir computing, was also performed with the IGR and good performance was achieved, with an NMSE of 0.163. A short-term memory task was performed to investigate an enhancement mechanism resulting from the IG addition. An increase in memory capacity, from 1.87 without IG to 2.73 with IG, was observed in the forgetting curves, indicating that enhancement of both high dimensionality and memory capacity are attributed to the origin of the performance improvement., Comment: 22 pages, 4 figures

Physical reservoir computing, which is a promising method for the implementation of highly efficient artificial intelligence devices, requires a physical system with nonlinearity, fading memory, and the ability to map in high dimensions. Although it is expected that spin wave interference can perform as highly efficient reservoir computing in some micromagnetic simulations, there has been no experimental verification to date. Herein, we demonstrate reservoir computing that utilizes multidetected nonlinear spin wave interference in an yttrium iron garnet single crystal. The subject computing system achieved excellent performance when used for hand-written digit recognition, second-order nonlinear dynamical tasks, and nonlinear autoregressive moving average (NARMA). It is of particular note that normalized mean square errors (NMSEs) for NARMA2 and second-order nonlinear dynamical tasks were 1.81x10-2 and 8.37x10-5, respectively, which are the lowest figures for any experimental physical reservoir so far reported. Said high performance was achieved with higher nonlinearity and the large memory capacity of interfered spin wave multi-detection., Comment: 19 pages, 5 figures

Physical reservoir computing has recently been attracting attention for its ability to significantly reduce the computational resources required to process time-series data. However, the physical reservoirs that have been reported to date have had insufficient expression power, and most of them have a large volume, which makes their practical application difficult. Herein we describe the development of a Li+-electrolyte based ion-gating reservoir (IGR), with ion-electron coupled dynamics, for use in high performance physical reservoir computing. A variety of synaptic responses were obtained in response to past experience, which responses were stored as transient charge density patterns in an electric double layer, at the Li+-electrolyte/diamond interface. Performance, which was tested using a nonlinear autoregressive moving-average (NARMA) task, was found to be excellent, with a NMSE of 0.023 for NARMA2, which is the highest for any physical reservoir reported to date. The maximum Lyapunov exponent of the IGR was 0.0083: the edge of chaos state enabling the best computational capacity. The IGR described herein opens the way for high-performance and integrated neural network devices., Comment: 25 pages, 5 figures for main part and 4 figures for Supplementary information