A memristor is a resistive device with an inherent memory. The theoretical concept of a memristor was connected to physically measured devices in 2008 and since then there has been rapid progress in the development of such devices, leading to a series of recent demonstrations of memristor-based neuromorphic hardware systems. Here, we evaluate the state of the art in memristor-based electronics and explore where the future of the field lies. We highlight three areas of potential technological impact: on-chip memory and storage, biologically inspired computing and general-purpose in-memory computing. We analyse the challenges, and possible solutions, associated with scaling the systems up for practical applications, and consider the benefits of scaling the devices down in terms of geometry and also in terms of obtaining fundamental control of the atomic-level dynamics. Finally, we discuss the ways we believe biology will continue to provide guiding principles for device innovation and system optimization in the field. This Perspective evaluates the state of the art in memristor-based electronics and explores the future development of such devices in on-chip memory, biologically inspired computing and general-purpose in-memory computing.

The future of electronics based on memristive systems

Memristor crossbars offer reconfigurable non-volatile resistance states and could remove the speed and energy efficiency bottleneck in vector-matrix multiplication, a core computing task in signal and image processing. Using such systems to multiply an analogue-voltage-amplitude-vector by an analogue-conductance-matrix at a reasonably large scale has, however, proved challenging due to difficulties in device engineering and array integration. Here we show that reconfigurable memristor crossbars composed of hafnium oxide memristors on top of metal-oxide-semiconductor transistors are capable of analogue vector-matrix multiplication with array sizes of up to 128 × 64 cells. Our output precision (5–8 bits, depending on the array size) is the result of high device yield (99.8%) and the multilevel, stable states of the memristors, while the linear device current–voltage characteristics and low wire resistance between cells leads to high accuracy. With the large memristor crossbars, we demonstrate signal processing, image compression and convolutional filtering, which are expected to be important applications in the development of the Internet of Things (IoT) and edge computing.

Analogue signal and image processing with large memristor crossbars

Memristors with tunable resistance states are emerging building blocks of artificial neural networks. However, in situ learning on a large-scale multiple-layer memristor network has yet to be demonstrated because of challenges in device property engineering and circuit integration. Here we monolithically integrate hafnium oxide-based memristors with a foundry-made transistor array into a multiple-layer neural network. We experimentally demonstrate in situ learning capability and achieve competitive classification accuracy on a standard machine learning dataset, which further confirms that the training algorithm allows the network to adapt to hardware imperfections. Our simulation using the experimental parameters suggests that a larger network would further increase the classification accuracy. The memristor neural network is a promising hardware platform for artificial intelligence with high speed-energy efficiency.

/pdf/efficient-and-self-adaptive-in-situ-learning-in-multilayer-1l25kt0b6s.pdf

Efficient and self-adaptive in-situ learning in multilayer memristor neural networks

The rapid increase in information in the big-data era calls for changes to information-processing paradigms, which, in turn, demand new circuit-building blocks to overcome the decreasing cost-effectiveness of transistor scaling and the intrinsic inefficiency of using transistors in non-von Neumann computing architectures. Accordingly, resistive switching materials (RSMs) based on different physical principles have emerged for memories that could enable energy-efficient and area-efficient in-memory computing. In this Review, we survey the four physical mechanisms that lead to such resistive switching: redox reactions, phase transitions, spin-polarized tunnelling and ferroelectric polarization. We discuss how these mechanisms equip RSMs with desirable properties for representation capability, switching speed and energy, reliability and device density. These properties are the key enablers of processing-in-memory platforms, with applications ranging from neuromorphic computing and general-purpose memcomputing to cybersecurity. Finally, we examine the device requirements for such systems based on RSMs and provide suggestions to address challenges in materials engineering, device optimization, system integration and algorithm design. Resistive switching materials enable novel, in-memory information processing, which may resolve the von Neumann bottleneck. This Review focuses on how the switching mechanisms and the resultant electrical properties lead to various computing applications.

Resistive switching materials for information processing

Sparse representation of information performs powerful feature extraction on high-dimensional data and is of interest for applications in signal processing, machine vision, object recognition, and neurobiology Sparse coding is a mechanism by which biological neural systems can efficiently process complex sensory data while consuming very little power Sparse coding algorithms in a bio-inspired approach can be implemented in a crossbar array of memristors (resistive memory devices) This network enables efficient implementation of pattern matching and lateral neuron inhibition, allowing input data to be sparsely encoded using neuron activities and stored dictionary elements The reconstructed input can be obtained by performing a backward pass through the same crossbar matrix using the neuron activity vector as input Different dictionary sets can be trained and stored in the same system, depending on the nature of the input signals Using the sparse coding algorithm, natural image processing is performed based on a learned dictionary

Sparse coding with Memristor networks

Time-series analysis including forecasting is essential in a range of fields from finance to engineering. However, long-term forecasting is difficult, particularly for cases where the underlying models and parameters are complex and unknown. Neural networks can effectively process features in temporal units and are attractive for such purposes. Reservoir computing, in particular, can offer efficient temporal processing of recurrent neural networks with a low training cost, and is thus well suited to time-series analysis and forecasting tasks. Here, we report a reservoir computing hardware system based on dynamic tungsten oxide (WOx) memristors that can efficiently process temporal data. The internal short-term memory effects of the WOx memristors allow the memristor-based reservoir to nonlinearly map temporal inputs into reservoir states, where the projected features can be readily processed by a linear readout function. We use the system to experimentally demonstrate two standard benchmarking tasks: isolated spoken-digit recognition with partial inputs, and chaotic system forecasting. A high classification accuracy of 99.2% is obtained for spoken-digit recognition, and autonomous chaotic time-series forecasting has been demonstrated over the long term. A reservoir computer system based on dynamic tungsten oxide memristors can be used to perform time-series analysis, demonstrating isolated spoken-digit recognition with partial inputs and chaotic system forecasting.

Temporal data classification and forecasting using a memristor-based reservoir computing system

A spin-polarized vertical-cavity surface-emitting laser is demonstrated with electrical spin injection from an $\mathrm{Fe}/{\mathrm{Al}}_{0.1}{\mathrm{Ga}}_{0.9}\mathrm{As}$ Schottky tunnel barrier. Laser operation with a spin-polarized current results in a maximum threshold current reduction of 11% and degree of circular polarization of 23% at 50 K. A cavity spin polarization of 16.8% is estimated from spin-dependent rate equation analysis of the observed threshold reduction.

Electrical spin injection and threshold reduction in a semiconductor laser.

Memristors have been considered as a leading candidate for a number of critical applications ranging from nonvolatile memory to non-Von Neumann computing systems. Feature extraction, which aims to transform input data from a high-dimensional space to a space with fewer dimensions, is an important technique widely used in machine learning and pattern recognition applications. Here, we experimentally demonstrate that memristor arrays can be used to perform principal component analysis, one of the most commonly used feature extraction techniques, through online, unsupervised learning. Using Sanger’s rule, that is, the generalized Hebbian algorithm, the principal components were obtained as the memristor conductances in the network after training. The network was then used to analyze sensory data from a standard breast cancer screening database with high classification success rate (97.1%).

Experimental Demonstration of Feature Extraction and Dimensionality Reduction Using Memristor Networks

Memristor-based neuromorphic networks have been actively studied as a promising candidate to overcome the von-Neumann bottleneck in future computing applications. Several recent studies have demonstrated memristor network’s capability to perform supervised as well as unsupervised learning, where features inherent in the input are identified and analyzed by comparing with features stored in the memristor network. However, even though in some cases the stored feature vectors can be normalized so that the winning neurons can be directly found by the (input) vector–(stored) vector dot-products, in many other cases, normalization of the feature vectors is not trivial or practically feasible, and calculation of the actual Euclidean distance between the input vector and the stored vector is required. Here we report experimental implementation of memristor crossbar hardware systems that can allow direct comparison of the Euclidean distances without normalizing the weights. The experimental system enables unsuperv...

K-means Data Clustering with Memristor Networks

The dependence of barrier height inhomogeneity on the gate metal has been investigated for the AlGaN/GaN Schottky diode. The analysis from the electroreflectance spectroscopy measurement for different types of Schottky gate metals tried (in this case, Au, Pt, Pd, and Ni) reveals that the surface donor states of AlGaN/GaN heterostructure strongly depends on the type of Schottky gate metals used, which suggests that barrier height inhomogeneity is strongly dependent on the gate metal. The X-ray photoelectron spectroscopy also reveals a strong correlation between the barrier height inhomogeneity and the gate metal type.

Jong Hoon Shin

Papers

Temporal data classification and forecasting using a memristor-based reservoir computing system

Electrical spin injection and threshold reduction in a semiconductor laser.

Experimental Demonstration of Feature Extraction and Dimensionality Reduction Using Memristor Networks

K-means Data Clustering with Memristor Networks

Metal induced inhomogeneous Schottky barrier height in AlGaN/GaN Schottky diode