Functional Materials for Memristor‐Based Reservoir Computing: Dynamics and Applications

Abstract

The booming development of artificial intelligence (AI) requires faster physical processing units as well as more efficient algorithms. Recently, reservoir computing (RC) has emerged as an alternative brain‐inspired framework for fast learning with low training cost, since only the weights associated with the output layers should be trained. Physical RC becomes one of the leading paradigms for computation using high‐dimensional, nonlinear, dynamic substrates. Among them, memristor appears to be a simple, adaptable, and efficient framework for constructing physical RC since they exhibit nonlinear features and memory behavior, while memristor‐implemented artificial neural networks display increasing popularity towards neuromorphic computing. In this review, the memristor‐implemented RC systems from the following aspects: architectures, materials, and applications are summarized. It starts with an introduction to the RC structures that can be simulated with memristor blocks. Specific interest then focuses on the dynamic memory behaviors of memristors based on various material systems, optimizing the understanding of the relationship between the relaxation behaviors and materials, which provides guidance and references for building RC systems coped with on‐demand application scenarios. Furthermore, recent advances in the application of memristor‐based physical RC systems are surveyed. In the end, the further prospects of memristor‐implemented RC system in a material view are envisaged.

Electron‐Sponge Nature of Polyoxometalates for Next‐Generation Electrocatalytic Water Splitting and Nonvolatile Neuromorphic Devices

“…Thus, the hybridization of active dopants (metal NPs and metal oxide NPs) with the organic matrix, modulating the charge injection barrier and ions transport properties, is a valuable strategy to control the CFs towards improving the electronic performance. [211] Figure 9c shows a supramolecular assembly strategy for POM clusters to realize charge trapping and CF formation in low-power NVMs for neuromorphic computing. Downscaling of integrated memories (PC main memory, USB memory, and SSD memory) to the sub-10 nm level [212] permits the integration of stimuli-responsive molecules, with 3D structural frameworks and discrete energy levels, as switching elements.…”

“…Therefore, unlike classical von Neumann computing, POM-based RC has displayed increasing popularity toward neuromorphic computing. [211]…”

“…Unfortunately, ANNs depend on the classical von Neumann computing which leads to unnecessary power consumption. [211] In contrast, neuromorphic computing, owing to the integration of highly parallelized computing architectures, enables intelligent systems to simulate neurobiology at the nanodevice level with lower power consumption. RC has been known as an alternative brain-inspired framework for fast learning with low training costs.…”

Multistate switching of scanning tunnelling microscopy machined polyoxovanadate–dysprosium–phthalocyanine nanopatterns on graphite

“…extremely small data memory elements that can change their logic state information by the potential-induced change in their resistive properties, 18,19 should have a positive impact on conventional computing as well as neuromorphic applications. 20,21 However, the immobilisation of POMs on various surfaces 22 and the revolutionary implementation of POM electronics in technical devices requires solving urgent problems of POM charge stabilisation on surfaces, multistate capacitance switching at room temperature (r.t.), spatially controlled positioning and nano-structuring of POMs, their crosstalk and electrical contacting, and, finally, efficient integration into existing production lines.…”

A Diffusive Artificial Synapse Based on Charged Metal Nanoparticles

“…This fast ion relaxation leads to good spike-dependent plasticity, especially at a high frequency (Figure ). Second, this device is set-step free and without a delay time, which is usually observed in threshold switching diffusive devices. , Consequently, it resembles more the Ca 2+ dynamics within biological synapses, leading to the good emulation of synaptic functions and the building of memristive operators for edge extraction with encouraging results. This feature also enables low-power operation, which is significantly important in memristive devices, especially in high-density arrays .…”