An efficient and highly performing memristor-based reservoir computing system

An efficient and high-performance memristor-based reservoir computing system

Figure summarizing the hardware architecture and application of the DM-RC system. Photo credit: Zhong et al.

Reservoir Computing (RC) is an approach to building computing systems inspired by current knowledge of the human brain. Neuromorphic computer architectures based on this approach consist of dynamic physical nodes that can process spatiotemporal signals in combination.

Researchers at Tsinghua University in China have recently developed a new RC system based on memristors, electrical components that regulate the flow of current in a circuit while also recording the amount of charge previously passed through it. This RC system featured in an article published in nature electronicshas been found to achieve remarkable results in both performance and efficiency.

“The basic architecture of our memristor RC system comes from our previous work published in nature communicationwhere we validated the feasibility of building an analog reservoir layer with dynamic memristors,” Jianshi Tang, one of the researchers who conducted the study, told TechXplore. “In this new work, we further build the analog readout layer with non-volatile memristors and integrate it with the dynamic memristor array-based parallel reservoir layer to implement a fully analog RC system.

The RC system developed by Tang and his colleagues is based on 24 dynamic memristors (DMs) connected to form a physical reservoir. Its readout layer, on the other hand, consists of 2048×4 non-volatile memristors (NVMs).

“Each DM in the DM-RC system is a physical system with computing power (referred to as a DM node) that can generate rich reservoir states through a time-division multiplexing process,” Tang explained. “These reservoir states are then loaded directly into the NVM array for multiply Accumulate (MAC) operations fed into the analog domain, resulting in the final output.”

Tang and his colleagues evaluated the performance of their dynamic memristor-based RC system by using it to run a deep learning model for two spatiotemporal signal processing tasks. They found that it achieved remarkably high classification accuracies of 96.6% and 97.9% for arrhythmia detection and dynamic gesture recognition, respectively.

“Compared to the digital RC system, our all-analog RC system has equivalent performance in accuracy but saves more than 99.9% of power consumption (22.2 μW vs. 29.4 mW),” Tang said unique feature of our work is that to build a fully analog RC system we used two different types of memristors: DMs as parallel reservoirs and NVM arrays as readout layer, without the help of digital components like those in previously reported hardware RC systems are used.”

The unique system architecture developed by this research team significantly reduces the complexity of RC approaches while significantly reducing power consumption. It could thus allow for simpler and larger RC hardware implementations in the future.

“Optimized non-volatile memristors with excellent analog switching characteristics were integrated to ensure consistent analog signal transmission and processing throughout the RC system,” Tang said. “Based on the noise model extracted from our memristor arrays, a noise-aware linear regression method was also used to train the output weight and effectively mitigate the loss of accuracy (less than 2%) caused by the non-ideal characteristics of memristors. “

Tang and his colleagues were the first to demonstrate real-time, all-analog signal processing using an RC hardware system. This demonstration finally allowed them to reliably evaluate the overall power consumption of their system.

“By correlating the experimental data with model simulations, the working mechanism of the DM-RC system, we were also able to find out more about the relationship between the electrical properties of physical nodes and the system performance,” Tang said. “More specifically, we revealed two key features ( i.e. threshold and window) extracted from dynamic memristor node properties that had a significant impact on reservoir quality.”

After identifying two characteristics that affected the performance of their RC system, Tang and his colleagues were able to define ranges of those two characteristics that led to optimal RC performance. Combined, these areas and their other findings could serve as a guide for future RC system design and optimization. This could help unlock their potential for edge computing along with other applications that require low power consumption and affordable hardware costs.

“In the future, the entire DM-RC system could be miniaturized and monolithically integrated on the chip to further reduce power consumption and computational latency,” Tang added. “Also, a deeper and more sophisticated RC system can be constructed using the DM-RC system as the base unit, which would further improve system performance due to richer reservoir states and stronger storage capacity.”

A reservoir computing system for temporal data classification and prediction

More information:
Yanan Zhong et al, A memristor-based analog reservoir computing system for real-time and low-power signal processing, nature electronics (2022). DOI: 10.1038/s41928-022-00838-3

Yanan Zhong et al., Dynamic memristor-based reservoir computing for high-efficiency temporal signal processing, nature communication (2021). DOI: 10.1038/s41467-020-20692-1

© 2022 Science X Network

Citation: An Efficient and Highly Performing Memristor-based Reservoir Computing System (2022 October 19) Retrieved October 19, 2022 from reservoir.html

This document is protected by copyright. Except for fair trade for the purpose of private study or research, no part may be reproduced without written permission. The content is for informational purposes only.