In materia reservoir computing with a fully memristive architecture based on self-organizing nanowire networks

Memristive Ion Dynamics to Enable Biorealistic Computing

Conventional artificial intelligence (AI) systems are facing bottlenecks due to the fundamental mismatches between AI models, which rely on parallel, in-memory, and dynamic computation, and traditional transistors, which have been designed and optimized for sequential logic operations. This calls for the development of novel computing units beyond transistors. Inspired by the high efficiency and adaptability of biological neural networks, computing systems mimicking the capabilities of biological structures are gaining more attention. Ion-based memristive devices (IMDs), owing to the intrinsic functional similarities to their biological counterparts, hold significant promise for implementing emerging neuromorphic learning and computing algorithms. In this article, we review the fundamental mechanisms of IMDs based on ion drift and diffusion to elucidate the origins of their diverse dynamics. We then examine how these mechanisms operate within different materials to enable IMDs with various types of switching behaviors, leading to a wide range of applications, from emulating biological components to realizing specialized computing requirements. Furthermore, we explore the potential for IMDs to be modified and tuned to achieve customized dynamics, which positions them as one of the most promising hardware candidates for executing bioinspired algorithms with unique specifications. Finally, we identify the challenges currently facing IMDs that hinder their widespread usage and highlight emerging research directions that could significantly benefit from incorporating IMDs.

Harnessing spatiotemporal transformation in magnetic domains for nonvolatile physical reservoir computing

Combining physics with computational models is increasingly recognized for enhancing the performance and energy efficiency in neural networks. Physical reservoir computing uses material dynamics of physical substrates for temporal data processing. Despite the ease of training, building an efficient reservoir remains challenging. Here, we explore beyond the conventional delay-based reservoirs by exploiting the spatiotemporal transformation in all-electric spintronic devices. Our nonvolatile spintronic reservoir effectively transforms the history dependence of reservoir states to the path dependence of domains. We configure devices triggered by different pulse widths as neurons, creating a reservoir featured by strong nonlinearity and rich interconnections. Using a small reservoir of merely 14 physical nodes, we achieved a high recognition rate of 0.903 in written digit recognition and a low error rate of 0.076 in Mackey-Glass time series prediction on a proof-of-concept printed circuit board. This work presents a promising route of nonvolatile physical reservoir computing, which is adaptable to the larger memristor family and broader physical neural networks.

Resistive memory-based zero-shot liquid state machine for multimodal event data learning

The human brain is a complex spiking neural network (SNN) capable of learning multimodal signals in a zero-shot manner by generalizing existing knowledge. Remarkably, it maintains minimal power consumption through event-based signal propagation. However, replicating the human brain in neuromorphic hardware presents both hardware and software challenges. Hardware limitations, such as the slowdown of Moore's law and Von Neumann bottleneck, hinder the efficiency of digital computers. In addition, SNNs are characterized by their software training complexities. Here, to this end, we propose a hardware-software co-design on a 40 nm 256 kB in-memory computing macro that physically integrates a fixed and random liquid state machine SNN encoder with trainable artificial neural network projections. We showcase the zero-shot learning of multimodal events on the N-MNIST and N-TIDIGITS datasets, including visual and audio data association, as well as neural and visual data alignment for brain-machine interfaces. Our co-design achieves classification accuracy comparable to fully optimized software models, resulting in a 152.83- and 393.07-fold reduction in training costs compared with state-of-the-art spiking recurrent neural network-based contrastive learning and prototypical networks, and a 23.34- and 160-fold improvement in energy efficiency compared with cutting-edge digital hardware, respectively. These proof-of-principle prototypes demonstrate zero-shot multimodal events learning capability for emerging efficient and compact neuromorphic hardware.

Weighted Echo State Graph Neural Networks Based on Robust and Epitaxial Film Memristors

Hardware system customized toward the demands of graph neural network learning would promote efficiency and strong temporal processing for graph-structured data. However, most amorphous/polycrystalline oxides-based memristors commonly have unstable conductance regulation due to random growth of conductive filaments. And graph neural networks based on robust and epitaxial film memristors can especially improve energy efficiency due to their high endurance and ultra-low power consumption. Here, robust and epitaxial Gd: HfO2-based film memristors are reported and construct a weighted echo state graph neural network (WESGNN). Benefiting from the optimized epitaxial films, the high switching speed (20 ns), low energy consumption (2.07 fJ), multi-value storage (4 bits), and high endurance (109) outperform most memristors. Notably, thanks to the appropriately dispersed conductance distribution (standard deviation = 7.68 nS), the WESGNN finely regulates the relative weights of input nodes and recursive matrix to realize state-of-the-art performance using the MUTAG and COLLAB datasets for graph classification tasks. Overall, robust and epitaxial film memristors offer nanoscale scalability, high reliability, and low energy consumption, making them energy-efficient hardware solutions for graph learning applications.

Hardware-Feasible and Efficient N-Type Organic Neuromorphic Signal Recognition via Reservoir Computing

Organic electrochemical synaptic transistors (OESTs), inspired by the biological nervous system, have garnered increasing attention due to their multifunctional applications in neuromorphic computing. However, the practical implementation of OESTs for signal recognition-particularly those utilizing n-type organic mixed ionic-electronic conductors (OMIECs)-still faces significant challenges at the hardware level. Here, a state-of-the-art small-molecule n-type OEST integrated within a physically simple and hardware feasible reservoir-computing (RC) framework for practical temporal signal recognition is presented. This integration is achieved by leveraging the adjustable synaptic properties of the n-OEST, which exhibits tunable nonlinear short-term memory, transitioning from volatility to nonvolatility, and demonstrating adaptive temporal specificity. Additionally, the nonvolatile OEST offers 256 conductance levels and a wide dynamic range (≈147) in long-term potentiation/depression (LTP/LTD), surpassing previously reported n-OESTs. By combining volatile n-OESTs as reservoirs with a single-layer perceptron readout composed of nonvolatile n-OEST networks, this physical RC system achieves substantial recognition accuracy for both handwritten-digit images (94.9%) and spoken digit (90.7%), along with ultrahigh weight efficiency. Furthermore, this system demonstrates outstanding accuracy (98.0%) by grouped RC in practical sleep monitoring, specifically in snoring recognition. Here, a reliable pathway for OMIEC-driven computing is presented to advance bioinspired hardware-based neuromorphic computing in the physical world.

Emerging 2D Ferroelectric Devices for In-Sensor and In-Memory Computing

The quantity of sensor nodes within current computing systems is rapidly increasing in tandem with the sensing data. The presence of a bottleneck in data transmission between the sensors, computing, and memory units obstructs the system's efficiency and speed. To minimize the latency of data transmission between units, novel in-memory and in-sensor computing architectures are proposed as alternatives to the conventional von Neumann architecture, aiming for data-intensive sensing and computing applications. The integration of 2D materials and 2D ferroelectric materials has been expected to build these novel sensing and computing architectures due to the dangling-bond-free surface, ultra-fast polarization flipping, and ultra-low power consumption of the 2D ferroelectrics. Here, the recent progress of 2D ferroelectric devices for in-sensing and in-memory neuromorphic computing is reviewed. Experimental and theoretical progresses on 2D ferroelectric devices, including passive ferroelectrics-integrated 2D devices and active ferroelectrics-integrated 2D devices, are reviewed followed by the integration of perception, memory, and computing application. Notably, 2D ferroelectric devices have been used to simulate synaptic weights, neuronal model functions, and neural networks for image processing. As an emerging device configuration, 2D ferroelectric devices have the potential to expand into the sensor-memory and computing integration application field, leading to new possibilities for modern electronics.

Iono-Magnonic Reservoir Computing With Chaotic Spin Wave Interference Manipulated by Ion-Gating

Physical reservoirs are a promising approach for realizing high-performance artificial intelligence devices utilizing physical devices. Although nonlinear interfered spin-wave multi-detection exhibits high nonlinearity and the ability to map in high dimensional feature space, it does not have sufficient performance to process time-series data precisely. Herein, development of an iono-magnonic reservoir by combining such interfered spin wave multi-detection and ion-gating involving protonation-induced redox reaction triggered by the application of voltage is reported. This study is the first to report the manipulation of the propagating spin wave property by ion-gating and the application of the same to physical reservoir computing. The subject iono-magnonic reservoir can generate various reservoir states in a single homogenous medium by utilizing a spin wave property modulated by ion-gating. Utilizing the strong nonlinearity resulting from chaos, the reservoir shows good computational performance in completing the Mackey-Glass chaotic time-series prediction task, and the performance is comparable to that exhibited by simulated neural networks.

Brain-like hardware, do we need it?

The brain's ability to perform efficient and fault-tolerant data processing is strongly related to its peculiar interconnected adaptive architecture, based on redundant neural circuits interacting at different scales. By emulating the brain's processing and learning mechanisms, computing technologies strive to achieve higher levels of energy efficiency and computational performance. Although efforts to address neuromorphic solutions through hardware based on top-down CMOS-based technologies have obtained interesting results in terms of energetic efficiency improvement, the replication of brain's self-assembled and redundant architectures is not considered in the roadmaps of data processing electronics. The exploration of solutions based on self-assembled elemental blocks to mimic biological networks' complexity is explored in the general frame of unconventional computing and it has not reached yet a maturity stage enabling a benchmark with standard electronic approaches in terms of performances, compatibility and scalability. Here we discuss some aspects related to advantages and disadvantages in the emulation of the brain for neuromorphic hardware. We also discuss possible directions in terms of hybrid hardware solutions where self-assembled substrates coexist and integrate with conventional electronics in view of neuromorphic architectures.

Bio-Plausible Multimodal Learning with Emerging Neuromorphic Devices

Multimodal machine learning, as a prospective advancement in artificial intelligence, endeavors to emulate the brain's multimodal learning abilities with the objective to enhance interactions with humans. However, this approach requires simultaneous processing of diverse types of data, leading to increased model complexity, longer training times, and higher energy consumption. Multimodal neuromorphic devices have the capability to preprocess spatio-temporal information from various physical signals into unified electrical signals with high information density, thereby enabling more biologically plausible multimodal learning with low complexity and high energy-efficiency. Here, this work conducts a comparison between the expression of multimodal machine learning and multimodal neuromorphic computing, followed by an overview of the key characteristics associated with multimodal neuromorphic devices. The bio-plausible operational principles and the multimodal learning abilities of emerging devices are examined, which are classified into heterogeneous and homogeneous multimodal neuromorphic devices. Subsequently, this work provides a detailed description of the multimodal learning capabilities demonstrated by neuromorphic circuits and their respective applications. Finally, this work highlights the limitations and challenges of multimodal neuromorphic computing in order to hopefully provide insight into potential future research directions.

Identifying winner-takes-all emergence in random nanowire networks: an inverse problem

Random nanowire networks (NWNs) are interconnects that enable the integration of nanoscopic building blocks (the nanowires) in a disorganized fashion, enabling the study of complex emergent phenomena in nanomaterials and built-in fault-tolerant processing functionalities; the latter can lead to advances in large-scale electronic devices that can be fabricated with no particular array/grid high-precision pattern. However, when various nanowires are assembled to form an intricate network, their individual features are somehow lost in the complex NWN frame, in line with the complexity hallmark "the whole differs from the sum of the parts". Individual nanowire materials and geometrical features can only be inferred indirectly by attempting to extract information about their initial conditions from a response function measurement. In this work, we present a mathematical framework that enables inference of the intrinsic properties of highly complex/intricate systems such as random NWNs in which information about their individual parts cannot be easily accessed due to their network formation and dynamical conductance behaviour falling in the category of memristive systems. Our method, named misfit minimization, is rooted in nonlinear regression supervised learning approaches in which we find the optimum parameters that minimize a cost function defined as the square least error between conductance evolution curves taken for a target NWN system and multiple configurational NWN samples composing the training set. The optimized parameters are features referent to the target NWN system's initial conditions obtained in an inverse fashion: from the response output function, we extract information about the target system's initial conditions. Accessing the nanowire individual features in a NWN frame, as our methodology allows, enables us to predict the conduction mechanisms of the NWN subjected to a current input source; these can be via a "winner-takes-all" energy-efficient scheme using a single conduction pathway composed of multiple nanowires connected in series or via multiple parallel conduction pathways. Predicting the conduction mechanism of complex and dynamical systems such as memristive NWNs is critical for their use in next-generation memory and brain-inspired technologies since their memory capability relies on the creation of such pathways activated and consolidated by the input current signal.

Nanomaterials for Flexible Neuromorphics

The quest to imbue machines with intelligence akin to that of humans, through the development of adaptable neuromorphic devices and the creation of artificial neural systems, has long stood as a pivotal goal in both scientific inquiry and industrial advancement. Recent advancements in flexible neuromorphic electronics primarily rely on nanomaterials and polymers owing to their inherent uniformity, superior mechanical and electrical capabilities, and versatile functionalities. However, this field is still in its nascent stage, necessitating continuous efforts in materials innovation and device/system design. Therefore, it is imperative to conduct an extensive and comprehensive analysis to summarize current progress. This review highlights the advancements and applications of flexible neuromorphics, involving inorganic nanomaterials (zero-/one-/two-dimensional, and heterostructure), carbon-based nanomaterials such as carbon nanotubes (CNTs) and graphene, and polymers. Additionally, a comprehensive comparison and summary of the structural compositions, design strategies, key performance, and significant applications of these devices are provided. Furthermore, the challenges and future directions pertaining to materials/devices/systems associated with flexible neuromorphics are also addressed. The aim of this review is to shed light on the rapidly growing field of flexible neuromorphics, attract experts from diverse disciplines (e.g., electronics, materials science, neurobiology), and foster further innovation for its accelerated development.

Advancements in Nanowire-Based Devices for Neuromorphic Computing: A Review

Neuromorphic computing, inspired by the highly interconnected and energy-efficient way the human brain processes information, has emerged as a promising technology for post-Moore's law era. This emerging technology can emulate the structures and the functions of the human brain and is expected to overcome the fundamental limitation of the current von Neumann computing architecture. Neuromorphic devices stand out as the key components of future electronic systems, exhibiting potential in shaping the landscape of neuromorphic computing. Especially, nanowire (NW)-based neuromorphic devices, with their advantages of high integration, high-speed computing, and low power consumption, have recently emerged as candidates for neuromorphic computing technology. Here, a critical overview of the current development and relevant research in the field of NW-based neuromorphic devices are provided. Neuromorphic devices based on different NW materials are comprehensively discussed, including Ag NW-based, organic NW-based, metal oxide NW-based, and semiconductor NW-based devices. Finally, as a foresight perspective, the potentials and the challenges of these NW-based neuromorphic devices for use as future brain-like electronics are discussed.

Parylene-MoOx crossbar memristors as a volatile reservoir and non-volatile readout: a homogeneous reservoir computing system

From the very beginning, the emulation of biological principles has been the primary avenue for the development of energy-efficient artificial intelligence systems. Reservoir computing, which has a solid biological basis, is particularly appealing due to its simplicity and efficiency. So-called memristors, resistive switching elements with complex dynamics, have proved beneficial for replicating both principal parts of a reservoir computing system. However, these parts require distinct behaviors found in differing memristive structures. The development of a homogeneous memristive reservoir computing system will significantly facilitate and reduce the fabrication process cost. The following work employs the co-existence of volatile and non-volatile regimes in parylene-MoOx crossbar memristors controlled by compliance current for this aim. The stable operation of the memristors under study is confirmed by low cycle-to-cycle and device-to-device variations of the switching voltages. For the transition between the volatile and non-volatile regimes, factors such as compliance current and reading voltage along with possible intrinsic origins are discussed. The results provide a foundation for the future hardware development of a homogeneous parylene-based reservoir computing system, considering high MNIST dataset classification accuracy (∼96%).

In-Situ Fabricated Transparent Flexible Nanowire Device with Wavelength-Regulated Dual-Function of Photodetector and Photonic Synapse

Integrating the dual functionalities of a photodetector and photonic synapse into a single device is challenging due to their conflicting requirements for photocurrent decay rates. This study addresses this issue by seamlessly depositing transparent indium tin oxide (ITO) electrodes onto self-oriented copper hexadecafluoro-phthalocyanine (F16CuPc) nanowires growing horizontally along hot-stamped periodic nanogrooves on a transparent flexible polyimide plastic film. This in-situ-fabricated device achieves bending-stable dual functionalities through wavelength regulation while maintaining high transparency and flexibility. Upon exposure to 450-850 nm light, the device exhibits a rapid and sensitive photoresponse with excellent bending stability, making it ideal for optical sensing in both visible and near-infrared spectra. More importantly, the device exhibits a bending-stable excitation postsynaptic current when exposed to light spikes below 405 nm. This enables the successful emulation of various biological synaptic functionalities, including paired-pulse facilitation, spike-number-dependent plasticity, spike-duration-dependent plasticity, spike-rating-dependent plasticity, configurable plasticity between short-term plasticity and long-term plasticity, and memory learning capabilities. Utilizing this device in an artificial neural network achieves a recognition rate of 95% after 57 training epochs. Its ability to switch between photodetection and synaptic modes by adjusting the light wavelength marks a significant advancement in the field of multifunctional flexible electronics based on nanowire arrays.

Analog reservoir computing via ferroelectric mixed phase boundary transistors

Analog reservoir computing (ARC) systems have attracted attention owing to their efficiency in processing temporal information. However, the distinct functionalities of the system components pose challenges for hardware implementation. Herein, we report a fully integrated ARC system that leverages material versatility of the ferroelectric-to-mixed phase boundary (MPB) hafnium zirconium oxides integrated onto indium-gallium-zinc oxide thin-film transistors (TFTs). MPB-based TFTs (MPBTFTs) with nonlinear short-term memory characteristics are utilized for physical reservoirs and artificial neuron, while nonvolatile ferroelectric TFTs mimic synaptic behavior for readout networks. Furthermore, double-gate configuration of MPBTFTs enhances reservoir state differentiation and state expansion for physical reservoir and processes both excitatory and inhibitory pulses for neuronal functionality with minimal hardware burden. The seamless integration of ARC components on a single wafer executes complex real-world time-series predictions with a low normalized root mean squared error of 0.28. The material-device co-optimization proposed in this study paves the way for the development of area- and energy-efficient ARC systems.

PEDOT:PSS Wire: A Two-Terminal Synaptic Device for Operation in Electrolyte and Saline Solutions

Synaptic devices, which are designed to emulate the synaptic functions of neurons, have recently gained attention as key elements in the development of neuromorphic hardware. To date, most synaptic devices utilizing conductive polymer materials, particularly poly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS), have been designed as three-terminal devices. Nevertheless, a recent study revealed that a single PEDOT:PSS wire can function as a two-terminal synaptic device through additional polymerization, which creates asymmetry in the wire diameter between the anode and cathode. Owing to its high biocompatibility, PEDOT is considered a promising candidate for use in clinical information-processing devices. However, previous studies examined the synaptic function of PEDOT:PSS only in PSS solutions. Therefore, the performance of PEDOT:PSS wires in other solutions, such as physiological saline solutions, remains unknown. Herein, we investigated the effects of operating environmental conditions (including phosphate-buffered saline (PBS)) on the synaptic functions of the asymmetric PEDOT:PSS wire. Our results indicate that the synaptic conductance change in the PEDOT:PSS wire occurred in all investigated aqueous electrolyte solutions. Moreover, we revealed the relationship between the synaptic conductance change behavior and the molecular properties of the electrolyte ions. Furthermore, the waveform of the conductance change can be controlled by adjusting the conditions for wire asymmetrization. These results demonstrate that the PEDOT:PSS wire exhibits a synaptic conductance change, yielding a waveform suitable for machine learning, even under wet conditions (i.e., in any electrolyte solution, including PBS). Therefore, PEDOT:PSS wire is a promising material for two-terminal synaptic devices applicable in clinical studies.

An overview of critical applications of resistive random access memory

The rapid advancement of new technologies has resulted in a surge of data, while conventional computers are nearing their computational limits. The prevalent von Neumann architecture, where processing and storage units operate independently, faces challenges such as data migration through buses, leading to decreased computing speed and increased energy loss. Ongoing research aims to enhance computing capabilities through the development of innovative chips and the adoption of new system architectures. One noteworthy advancement is Resistive Random Access Memory (RRAM), an emerging memory technology. RRAM can alter its resistance through electrical signals at both ends, retaining its state even after power-down. This technology holds promise in various areas, including logic computing, neural networks, brain-like computing, and integrated technologies combining sensing, storage, and computing. These cutting-edge technologies offer the potential to overcome the performance limitations of traditional architectures, significantly boosting computing power. This discussion explores the physical mechanisms, device structure, performance characteristics, and applications of RRAM devices. Additionally, we delve into the potential future adoption of these technologies at an industrial scale, along with prospects and upcoming research directions.

Toward a Brain-Neuromorphics Interface

Brain-computer interfaces (BCIs) that enable human-machine interaction have immense potential in restoring or augmenting human capabilities. Traditional BCIs are realized based on complementary metal-oxide-semiconductor (CMOS) technologies with complex, bulky, and low biocompatible circuits, and suffer with the low energy efficiency of the von Neumann architecture. The brain-neuromorphics interface (BNI) would offer a promising solution to advance the BCI technologies and shape the interactions with machineries. Neuromorphic devices and systems are able to provide substantial computation power with extremely high energy-efficiency by implementing in-materia computing such as in situ vector-matrix multiplication (VMM) and physical reservoir computing. Recent progresses on integrating neuromorphic components with sensing and/or actuating modules, give birth to the neuromorphic afferent nerve, efferent nerve, sensorimotor loop, and so on, which has advanced the technologies for future neurorobotics by achieving sophisticated sensorimotor capabilities as the biological system. With the development on the compact artificial spiking neuron and bioelectronic interfaces, the seamless communication between a BNI and a bioentity is reasonably expectable. In this review, the upcoming BNIs are profiled by introducing the brief history of neuromorphics, reviewing the recent progresses on related areas, and discussing the future advances and challenges that lie ahead.

Neuromorphic Nanoionics for Human-Machine Interaction: From Materials to Applications

Human-machine interaction (HMI) technology has undergone significant advancements in recent years, enabling seamless communication between humans and machines. Its expansion has extended into various emerging domains, including human healthcare, machine perception, and biointerfaces, thereby magnifying the demand for advanced intelligent technologies. Neuromorphic computing, a paradigm rooted in nanoionic devices that emulate the operations and architecture of the human brain, has emerged as a powerful tool for highly efficient information processing. This paper delivers a comprehensive review of recent developments in nanoionic device-based neuromorphic computing technologies and their pivotal role in shaping the next-generation of HMI. Through a detailed examination of fundamental mechanisms and behaviors, the paper explores the ability of nanoionic memristors and ion-gated transistors to emulate the intricate functions of neurons and synapses. Crucial performance metrics, such as reliability, energy efficiency, flexibility, and biocompatibility, are rigorously evaluated. Potential applications, challenges, and opportunities of using the neuromorphic computing technologies in emerging HMI technologies, are discussed and outlooked, shedding light on the fusion of humans with machines.

Neuromorphic overparameterisation and few-shot learning in multilayer physical neural networks

Physical neuromorphic computing, exploiting the complex dynamics of physical systems, has seen rapid advancements in sophistication and performance. Physical reservoir computing, a subset of neuromorphic computing, faces limitations due to its reliance on single systems. This constrains output dimensionality and dynamic range, limiting performance to a narrow range of tasks. Here, we engineer a suite of nanomagnetic array physical reservoirs and interconnect them in parallel and series to create a multilayer neural network architecture. The output of one reservoir is recorded, scaled and virtually fed as input to the next reservoir. This networked approach increases output dimensionality, internal dynamics and computational performance. We demonstrate that a physical neuromorphic system can achieve an overparameterised state, facilitating meta-learning on small training sets and yielding strong performance across a wide range of tasks. Our approach's efficacy is further demonstrated through few-shot learning, where the system rapidly adapts to new tasks.

Unique Coexistence of Two Resistive Switching Modes in a Memristor Device Enables Multifunctional Neuromorphic Computing Properties

We report on hybrid memristor devices consisting of germanium dioxide nanoparticles (GeO2 NP) embedded within a poly(methyl methacrylate) (PMMA) thin film. Besides exhibiting forming-free resistive switching and an uncommon "ON" state in pristine conditions, the hybrid (nanocomposite) devices demonstrate a unique form of mixed-mode switching. The observed stopping voltage-dependent switching enables state-of-the-art bifunctional synaptic behavior with short-term (volatile/temporal) and long-term (nonvolatile/nontemporal) modes that are switchable depending on the stopping voltage applied. The short-term memory mode device is demonstrated to further emulate important synaptic functions such as short-term potentiation (STP), short-term depression (STD), paired-pulse facilitation (PPF), post-tetanic potentiation (PTP), spike-voltage-dependent plasticity (SVDP), spike-duration-dependent plasticity (SDDP), and, more importantly, the "learning-forgetting-rehearsal" behavior. The long-term memory mode gives additional long-term potentiation (LTP) and long-term depression (LTD) characteristics for long-term plasticity applications. The work shows a unique coexistence of the two resistive switching modes, providing greater flexibility in device design for future adaptive and reconfigurable neuromorphic computing systems at the hardware level.

Semantic memory-based dynamic neural network using memristive ternary CIM and CAM for 2D and 3D vision

The brain is dynamic, associative, and efficient. It reconfigures by associating the inputs with past experiences, with fused memory and processing. In contrast, AI models are static, unable to associate inputs with past experiences, and run on digital computers with physically separated memory and processing. We propose a hardware-software co-design, a semantic memory-based dynamic neural network using a memristor. The network associates incoming data with the past experience stored as semantic vectors. The network and the semantic memory are physically implemented on noise-robust ternary memristor-based computing-in-memory (CIM) and content-addressable memory (CAM) circuits, respectively. We validate our co-designs, using a 40-nm memristor macro, on ResNet and PointNet++ for classifying images and three-dimensional points from the MNIST and ModelNet datasets, which achieves not only accuracy on par with software but also a 48.1 and 15.9% reduction in computational budget. Moreover, it delivers a 77.6 and 93.3% reduction in energy consumption.

High-Dimensional Physical Reservoir with Back-End-of-Line-Compatible Tin Monoxide Thin-Film Transistor

This work demonstrates a physical reservoir using a back-end-of-line compatible thin-film transistor (TFT) with tin monoxide (SnO) as the channel material for neuromorphic computing. The electron trapping and time-dependent detrapping at the channel interface induce the SnO·TFT to exhibit fading memory and nonlinearity characteristics, the critical assets for physical reservoir computing. The three-terminal configuration of the TFT allows the generation of higher-dimensional reservoir states by simultaneously adjusting the bias conditions of the gate and drain terminals, surpassing the performances of typical two-terminal-based reservoirs such as memristors. The high-dimensional SnO TFT reservoir performs exceptionally in two benchmark tests, achieving a 94.1% accuracy in Modified National Institute of Standards and Technology handwritten number recognition and a normalized root-mean-square error of 0.089 in Mackey-Glass time-series prediction. Furthermore, it is suitable for vertical integration because its fabrication temperature is <250 °C, providing the benefit of achieving a high integration density.

SiC@NiO Core-Shell Nanowire Networks-Based Optoelectronic Synapses for Neuromorphic Computing and Visual Systems at High Temperature

1D nanowire networks, sharing similarities of structure, information transfer, and computation with biological neural networks, have emerged as a promising platform for neuromorphic systems. Based on brain-like structures of 1D nanowire networks, neuromorphic synaptic devices can overcome the von Neumann bottleneck, achieving intelligent high-efficient sensing and computing function with high information processing rates and low power consumption. Here, high-temperature neuromorphic synaptic devices based on SiC@NiO core-shell nanowire networks optoelectronic memristors (NNOMs) are developed. Experimental results demonstrate that NNOMs attain synaptic short/long-term plasticity and modulation plasticity under both electrical and optical stimulation, and exhibit advanced functions such as short/long-term memory and "learning-forgetting-relearning" under optical stimulation at both room temperature and 200 °C. Based on the advanced functions under light stimulus, the constructed 5 × 3 optoelectronic synaptic array devices exhibit a stable visual memory function up to 200 °C, which can be utilized to develop artificial visual systems. Additionally, when exposed to multiple electronic or optical stimuli, the NNOMs effectively replicate the principles of Pavlovian classical conditioning, achieving visual heterologous synaptic functionality and refining neural networks. Overall, with abundant synaptic characteristics and high-temperature thermal stability, these neuromorphic synaptic devices offer a promising route for advancing neuromorphic computing and visual systems.

Improvement of volatile switching in scaled silicon nanofin memristor for high performance and efficient reservoir computing

Conductive-bridge random access memory can be used as a physical reservoir for temporal learning in reservoir computing owing to its volatile nature. Herein, a scaled Cu/HfOx/n+-Si memristor was fabricated and characterized for reservoir computing. The scaled, silicon nanofin bottom electrode formation is verified by scanning electron and transmission electron microscopy. The scaled device shows better cycle-to-cycle switching variability characteristics compared with those of large-sized cells. In addition, synaptic characteristics such as conductance changes due to pulses, paired-pulse facilitation, and excitatory postsynaptic currents are confirmed in the scaled memristor. High-pattern accuracy is demonstrated by deep neural networks applied in neuromorphic systems in conjunction with the use of the Modified National Institute of Standards and Technology database. Furthermore, a reservoir computing system is introduced with six different states attained by adjusting the amplitude of the input pulse. Finally, high-performance and efficient volatile reservoir computing in the scaled device is demonstrated by conductance control and system-level reservoir computing simulations.

Memristors with Tunable Volatility for Reconfigurable Neuromorphic Computing

Neuromorphic computing promises an energy-efficient alternative to traditional digital processors in handling data-heavy tasks, primarily driven by the development of both volatile (neuronal) and nonvolatile (synaptic) resistive switches or memristors. However, despite their energy efficiency, memristor-based technologies presently lack functional tunability, thus limiting their competitiveness with arbitrarily programmable (general purpose) digital computers. This work introduces a two-terminal bilayer memristor, which can be tuned among neuronal, synaptic, and hybrid behaviors. The varying behaviors are accessed via facile control over the filament formed within the memristor, enabled by the interplay between the two active ionic species (oxygen vacancies and metal cations). This solution is unlike single-species ion migration employed in most other memristors, which makes their behavior difficult to control. By reconfiguring a single crossbar array of hybrid memristors, two different applications that usually require distinct types of devices are demonstrated - reprogrammable heterogeneous reservoir computing and arbitrary non-Euclidean graph networks. Thus, this work outlines a potential path toward functionally reconfigurable postdigital computers.

Chemical reservoir computation in a self-organizing reaction network

Chemical reaction networks, such as those found in metabolism and signalling pathways, enable cells to process information from their environment1,2. Current approaches to molecular information processing and computation typically pursue digital computation models and require extensive molecular-level engineering3. Despite considerable advances, these approaches have not reached the level of information processing capabilities seen in living systems. Here we report on the discovery and implementation of a chemical reservoir computer based on the formose reaction4. We demonstrate how this complex, self-organizing chemical reaction network can perform several nonlinear classification tasks in parallel, predict the dynamics of other complex systems and achieve time-series forecasting. This in chemico information processing system provides proof of principle for the emergent computational capabilities of complex chemical reaction networks, paving the way for a new class of biomimetic information processing systems.

Experimental Demonstration of Reservoir Computing with Self-Assembled Percolating Networks of Nanoparticles

The complex self-assembled network of neurons and synapses that comprises the biological brain enables natural information processing with remarkable efficiency. Percolating networks of nanoparticles (PNNs) are complex self-assembled nanoscale systems that have been shown to possess many promising brain-like attributes and which are therefore appealing systems for neuromorphic computation. Here experiments are performed that show that PNNs can be utilized as physical reservoirs within a nanoelectronic reservoir computing framework and demonstrate successful computation for several benchmark tasks (chaotic time series prediction, nonlinear transformation, and memory capacity). For each task, relevant literature results are compiled and it is shown that the performance of the PNNs compares favorably to that previously reported from nanoelectronic reservoirs. It is then demonstrated experimentally that PNNs can be used for spoken digit recognition with state-of-the-art accuracy. Finally, a parallel reservoir architecture is emulated, which increases the dimensionality and richness of the reservoir outputs and results in further improvements in performance across all tasks.

2Memristor-1Capacitor Integrated Temporal Kernel for High-Dimensional Data Mapping

Compact but precise feature-extracting ability is core to processing complex computational tasks in neuromorphic hardware. Physical reservoir computing (RC) offers a robust framework to map temporal data into a high-dimensional space using the time dynamics of a material system, such as a volatile memristor. However, conventional physical RC systems have limited dynamics for the given material properties, restricting the methods to increase their dimensionality. This study proposes an integrated temporal kernel composed of a 2-memristor and 1-capacitor (2M1C) using a W/HfO2/TiN memristor and TiN/ZrO2/Al2O3/ZrO2/TiN capacitor to achieve higher dimensionality and tunable dynamics. The kernel elements are carefully designed and fabricated into an integrated array, of which performances are evaluated under diverse conditions. By optimizing the time dynamics of the 2M1C kernel, each memristor simultaneously extracts complementary information from input signals. The MNIST benchmark digit classification task achieves a high accuracy of 94.3% with a (196×10) single-layer network. Analog input mapping ability is tested with a Mackey-Glass time series prediction, and the system records a normalized root mean square error of 0.04 with a 20×1 readout network, the smallest readout network ever used for Mackey-Glass prediction in RC. These performances demonstrate its high potential for efficient temporal data analysis.

Highly-integrable analogue reservoir circuits based on a simple cycle architecture

Physical reservoir computing is a promising solution for accelerating artificial intelligence (AI) computations. Various physical systems that exhibit nonlinear and fading-memory properties have been proposed as physical reservoirs. Highly-integrable physical reservoirs, particularly for edge AI computing, has a strong demand. However, realizing a practical physical reservoir with high performance and integrability remains challenging. Herein, we present an analogue circuit reservoir with a simple cycle architecture suitable for complementary metal-oxide-semiconductor (CMOS) chip integration. In several benchmarks and demonstrations using synthetic and real-world data, our developed hardware prototype and its simulator exhibit a high prediction performance and sufficient memory capacity for practical applications, showing promise for future applications in highly integrated AI accelerators.

Reservoir Computing Based on Oxygen-Vacancy-Mediated X-ray Optical Synaptic Device for Medical CT Bone Diagnosis

Recognition and judgment of X-ray computed tomography (CT) images play a crucial role in medical diagnosis and disease prevention. However, the storage and calculation of the X-ray imaging system applied in the traditional CT diagnosis is separate, and the pathological judgment is based on doctors' experience, which will affect the timeliness and accuracy of decision-making. In this paper, a simple-structured reservoir computing network (RC) is proposed based on Ga2O3 X-ray optical synaptic devices to recognize medical skeletal CT images with high accuracy. Through oxygen vacancy engineering, Ga2O3 X-ray optical synaptic devices with adjustable photocurrent gain and a persistent photoconductivity effect were obtained. By using the Ga2O3 X-ray optical synaptic device as a reservoir, we constructed an RC network for medical skeletal CT diagnosis and verified its image recognition capability using the MNIST data set with an accuracy of 78.08%. In the elbow skeletal CT image recognition task, the recognition rate is as high as 100%. This work constructs a simple-structured RC network for X-ray image recognition, which is of great significance in applications in medical fields.

All-Photolithography Fabrication of Ion-Gated Flexible Organic Transistor Array for Multimode Neuromorphic Computing

Organic ion-gated transistors (OIGTs) demonstrate commendable performance for versatile neuromorphic systems. However, due to the fragility of organic materials to organic solvents, efficient and reliable all-photolithography methods for scalable manufacturing of high-density OIGT arrays with multimode neuromorphic functions are still missing, especially when all active layers are patterned in high-density. Here, a flexible high-density (9662 devices per cm2) OIGT array with high yield and minimal device-to-device variation is fabricated by a modified all-photolithography method. The unencapsulated flexible array can withstand 1000 times' bending at a radius of 1 mm, and 3 months' storage test in air, without obvious performance degradation. More interesting, the OIGTs can be configured between volatile and nonvolatile modes, suitable for constructing reservoir computing systems to achieve high accuracy in classifying handwritten digits with low training costs. This work proposes a promising design of organic and flexible electronics for affordable neuromorphic systems, encompassing both array and algorithm aspects.

Reconfigurable optoelectronic transistors for multimodal recognition

Biological nervous system outperforms in both dynamic and static information perception due to their capability to integrate the sensing, memory and processing functions. Reconfigurable neuromorphic transistors, which can be used to emulate different types of biological analogues in a single device, are important for creating compact and efficient neuromorphic computing networks, but their design remains challenging due to the need for opposing physical mechanisms to achieve different functions. Here we report a neuromorphic electrolyte-gated transistor that can be reconfigured to perform physical reservoir and synaptic functions. The device exhibits dynamics with tunable time-scales under optical and electrical stimuli. The nonlinear volatile property is suitable for reservoir computing, which can be used for multimodal pre-processing. The nonvolatility and programmability of the device through ion insertion/extraction achieved via electrolyte gating, which are required to realize synaptic functions, are verified. The device's superior performance in mimicking human perception of dynamic and static multisensory information based on the reconfigurable neuromorphic functions is also demonstrated. The present study provides an exciting paradigm for the realization of multimodal reconfigurable devices and opens an avenue for mimicking biological multisensory fusion.

Inkjet printed IGZO memristors with volatile and non-volatile switching

Solution-based memristors deposited by inkjet printing technique have a strong technological potential based on their scalability, low cost, environmentally friendlier processing by being an efficient technique with minimal material waste. Indium-gallium-zinc oxide (IGZO), an oxide semiconductor material, shows promising resistive switching properties. In this work, a printed Ag/IGZO/ITO memristor has been fabricated. The IGZO thickness influences both memory window and switching voltage of the devices. The devices show both volatile counter8wise (c8w) and non-volatile 8wise (8w) switching at low operating voltage. The 8w switching has a SET and RESET voltage lower than 2 V and - 5 V, respectively, a retention up to 105 s and a memory window up to 100, whereas the c8w switching shows volatile characteristics with a low threshold voltage (Vth < - 0.65 V) and a characteristic time (τ) of 0.75 ± 0.12 ms when a single pulse of - 0.65 V with width of 0.1 ms is applied. The characteristic time alters depending on the number of pulses. These volatile characteristics allowed them to be tested on different 4-bit pulse sequences, as an initial proof of concept for temporal signal processing applications.

3D-integrated multilayered physical reservoir array for learning and forecasting time-series information

A wide reservoir computing system is an advanced architecture composed of multiple reservoir layers in parallel, which enables more complex and diverse internal dynamics for multiple time-series information processing. However, its hardware implementation has not yet been realized due to the lack of a high-performance physical reservoir and the complexity of fabricating multiple stacks. Here, we achieve a proof-of-principle demonstration of such hardware made of a multilayered three-dimensional stacked 3 × 10 × 10 tungsten oxide memristive crossbar array, with which we further realize a wide physical reservoir computing for efficient learning and forecasting of multiple time-series data. Because a three-layer structure allows the seamless and effective extraction of intricate three-dimensional local features produced by various temporal inputs, it can readily outperform two-dimensional based approaches extensively studied previously. Our demonstration paves the way for wide physical reservoir computing systems capable of efficiently processing multiple dynamic time-series information.

Few- and single-molecule reservoir computing experimentally demonstrated with surface-enhanced Raman scattering and ion gating

Molecule-based reservoir computing (RC) is promising for achieving low power consumption neuromorphic computing, although the information-processing capability of small numbers of molecules is not clear. Here, we report a few- and single-molecule RC that uses the molecular vibration dynamics in the para-mercaptobenzoic acid (pMBA) detected by surface-enhanced Raman scattering (SERS) with tungsten oxide nanorod/silver nanoparticles. The Raman signals of the pMBA molecules, adsorbed at the SERS active site of the nanorod, were reversibly perturbated by the application of voltage-induced local pH changes near the molecules, and then used to perform time-series analysis tasks. Despite the small number of molecules used, our system achieved good performance, including >95% accuracy in various nonlinear waveform transformations, 94.3% accuracy in solving a second-order nonlinear dynamic system, and a prediction error of 25.0 milligrams per deciliter in a 15-minute-ahead blood glucose level prediction. Our work provides a concept of few-molecular computing with practical computation capabilities.

Silver Nanowire Networks with Moisture-Enhanced Learning Ability

The human brain possesses a remarkable ability to memorize information with the assistance of a specific external environment. Therefore, mimicking the human brain's environment-enhanced learning abilities in artificial electronic devices is essential but remains a considerable challenge. Here, a network of Ag nanowires with a moisture-enhanced learning ability, which can mimic long-term potentiation (LTP) synaptic plasticity at an ultralow operating voltage as low as 0.01 V, is presented. To realize a moisture-enhanced learning ability and to adjust the aggregations of Ag ions, we introduced a thin polyvinylpyrrolidone (PVP) coating layer with moisture-sensitive properties to the surfaces of the Ag nanowires of Ag ions. That Ag nanowire network was shown to exhibit, in response to the humidity of its operating environment, different learning speeds during the LTP process. In high-humidity environments, the synaptic plasticity was significantly strengthened with a higher learning speed compared with that in relatively low-humidity environments. Based on experimental and simulation results, we attribute this enhancement to the higher electric mobility of the Ag ions in the water-absorbed PVP layer. Finally, we demonstrated by simulation that the moisture-enhanced synaptic plasticity enabled the device to adjust connection weights and delivery modes based on various input patterns. The recognition rate of a handwritten data set reached 94.5% with fewer epochs in a high-humidity environment. This work shows the feasibility of building our electronic device to achieve artificial adaptive learning abilities.

Electrochemical rewiring through quantum conductance effects in single metallic memristive nanowires

Memristive devices have been demonstrated to exhibit quantum conductance effects at room temperature. In these devices, a detailed understanding of the relationship between electrochemical processes and ionic dynamic underlying the formation of atomic-sized conductive filaments and corresponding electronic transport properties in the quantum regime still represents a challenge. In this work, we report on quantum conductance effects in single memristive Ag nanowires (NWs) through a combined experimental and simulation approach that combines advanced classical molecular dynamics (MD) algorithms and quantum transport simulations (DFT). This approach provides new insights on quantum conductance effects in memristive devices by unravelling the intrinsic relationship between electronic transport and atomic dynamic reconfiguration of the nanofilment, by shedding light on deviations from integer multiples of the fundamental quantum of conductance depending on peculiar dynamic trajectories of nanofilament reconfiguration and on conductance fluctuations relying on atomic rearrangement due to thermal fluctuations.

Thin-film transistor for temporal self-adaptive reservoir computing with closed-loop architecture

Reservoir computing is a powerful neural network-based computing paradigm for spatiotemporal signal processing. Recently, physical reservoirs have been explored based on various electronic devices with outstanding efficiency. However, the inflexible temporal dynamics of these reservoirs have posed fundamental restrictions in processing spatiotemporal signals with various timescales. Here, we fabricated thin-film transistors with controllable temporal dynamics, which can be easily tuned with electrical operation signals and showed excellent cycle-to-cycle uniformity. Based on this, we constructed a temporal adaptive reservoir capable of extracting temporal information of multiple timescales, thereby achieving improved accuracy in the human-activity-recognition task. Moreover, by leveraging the former computing output to modify the hyperparameters, we constructed a closed-loop architecture that equips the reservoir computing system with temporal self-adaptability according to the current input. The adaptability is demonstrated by accurate real-time recognition of objects moving at diverse speed levels. This work provides an approach for reservoir computing systems to achieve real-time processing of spatiotemporal signals with compound temporal characteristics.

Unveiling transient current response in bilayer oxide-based physical reservoirs for time-series data analysis

Physical reservoirs employed to map time-series data and analyze extracted features have attracted interest owing to their low training cost and mitigated interconnection complexity. This study reports a physical reservoir based on a bilayer oxide-based dynamic memristor. The proposed device exhibits a nonlinear current response and short-term memory (STM), satisfying the requirements of reservoir computing (RC). These characteristics are validated using a compact model to account for resistive switching (RS) via the dynamic evolution of the internal state variable and the relocation of oxygen vacancies. Mathematically, the transient current response can be quantitatively described according to a simple set of equations to correlate the theoretical framework with experimental results. Furthermore, the device shows significant reliability and ability to distinguish 4-bit inputs and four diverse neural firing patterns. Therefore, this work shows the feasibility of implementing physical reservoirs in hardware and advances the understanding of the dynamic response.

Nonlinear Boost of Optical Angular Momentum Selectivity by Hybrid Nanolaser Circuits

Selective control of light is essential for optical science and technology, with numerous applications. However, optical selectivity in the angular momentum of light has been quite limited, remaining constant by increasing the incident light power on previous passive optical devices. Here, we demonstrate a nonlinear boost of optical selectivity in both the spin and orbital angular momentum of light through near-field selective excitation of single-mode nanolasers. Our designed hybrid nanolaser circuits consist of plasmonic metasurfaces and individually placed perovskite nanowires, enabling subwavelength focusing of angular-momentum-distinctive plasmonic fields and further selective excitation of nanolasers in nanowires. The optically selected nanolaser with a nonlinear increase of light emission greatly enhances the baseline optical selectivity offered by the metasurface from about 0.4 up to near unity. Our demonstrated hybrid nanophotonic platform may find important applications in all-optical logic gates and nanowire networks, ultrafast optical switches, nanophotonic detectors, and on-chip optical and quantum information processing.

Spatiotemporal Data Processing with Memristor Crossbar-Array-Based Graph Reservoir

Memristor-based physical reservoir computing (RC) is a robust framework for processing complex spatiotemporal data parallelly. However, conventional memristor-based reservoirs cannot capture the spatial relationship between the time-varying inputs due to the specific mapping scheme assigning one input signal to one memristor conductance. Here, a physical "graph reservoir" is introduced using a metal cell at the diagonal-crossbar array (mCBA) with dynamic self-rectifying memristors. Input and inverted input signals are applied to the word and bit lines of the mCBA, respectively, storing the correlation information between input signals in the memristors. In this way, the mCBA graph reservoirs can map the spatiotemporal correlation of the input data in a high-dimensional feature space. The high-dimensional mapping characteristics of the graph reservoir achieve notable results, including a normalized root-mean-square error of 0.09 in Mackey-Glass time series prediction, a 97.21% accuracy in MNIST recognition, and an 80.0% diagnostic accuracy in human connectome classification.

Adapted MLP-Mixer network based on crossbar arrays of fast and multilevel switching (Co-Fe-B)x(LiNbO3)100-x nanocomposite memristors

MLP-Mixer based on multilayer perceptrons (MLPs) is a novel architecture of a neuromorphic computing system (NCS) introduced for image classification tasks without convolutional layers. Its software realization demonstrates high classification accuracy, although the number of trainable weights is relatively low. One more promising way of improving the NCS performance, especially in terms of power consumption, is its hardware realization using memristors. Therefore, in this work, we proposed an NCS with an adapted MLP-Mixer architecture and memristive weights. For this purpose, we used a passive crossbar array of (Co-Fe-B)x(LiNbO3)100-x memristors. Firstly, we studied the characteristics of such memristors, including their minimal resistive switching time, which was extrapolated to be in the picosecond range. Secondly, we created a fully hardware NCS with memristive weights that are capable of classification of simple 4-bit vectors. The system was shown to be robust to noise introduction in the input patterns. Finally, we used experimental memristive characteristics to simulate an adapted MLP-Mixer architecture that demonstrated a classification accuracy of (94.7 ± 0.3)% on the Modified National Institute of Standards and Technology (MNIST) dataset. The obtained results are the first steps toward the realization of memristive NCS with a promising MLP-Mixer architecture.

Memristor-Based Artificial Chips

Memristors, promising nanoelectronic devices with in-memory resistive switching behavior that is assembled with a physically integrated core processing unit (CPU) and memory unit and even possesses highly possible multistate electrical behavior, could avoid the von Neumann bottleneck of traditional computing devices and show a highly efficient ability of parallel computation and high information storage. These advantages position them as potential candidates for future data-centric computing requirements and add remarkable vigor to the research of next-generation artificial intelligence (AI) systems, particularly those that involve brain-like intelligence applications. This work provides an overview of the evolution of memristor-based devices, from their initial use in creating artificial synapses and neural networks to their application in developing advanced AI systems and brain-like chips. It offers a broad perspective of the key device primitives enabling their special applications from the view of materials, nanostructure, and mechanism models. We highlight these demonstrations of memristor-based nanoelectronic devices that have potential for use in the field of brain-like AI, point out the existing challenges of memristor-based nanodevices toward brain-like chips, and propose the guiding principle and promising outlook for future device promotion and system optimization in the biomedical AI field.

Integrated In-Memory Sensor and Computing of Artificial Vision Based on Full-vdW Optoelectronic Ferroelectric Field-Effect Transistor

The development and application of artificial intelligence have led to the exploitation of low-power and compact intelligent information-processing systems integrated with sensing, memory, and neuromorphic computing functions. The 2D van der Waals (vdW) materials with abundant reservoirs for arbitrary stacking based on functions and enabling continued device downscaling offer an attractive alternative for continuously promoting artificial intelligence. In this study, full 2D SnS2 /h-BN/CuInP2 S6 (CIPS)-based ferroelectric field-effect transistors (Fe-FETs) and utilized light-induced ferroelectric polarization reversal to achieve excellent memory properties and multi-functional sensing-memory-computing vision simulations are designed. The device exhibits a high on/off current ratio of over 105 , long retention time (>104  s), stable cyclic endurance (>350 cycles), and 128 multilevel current states (7-bit). In addition, fundamental synaptic plasticity characteristics are emulated including paired-pulse facilitation (PPF), short-term plasticity (STP), long-term plasticity (LTP), long-term potentiation, and long-term depression. A ferroelectric optoelectronic reservoir computing system for the Modified National Institute of Standards and Technology (MNIST) handwritten digital recognition achieved a high accuracy of 93.62%. Furthermore, retina-like light adaptation and Pavlovian conditioning are successfully mimicked. These results provide a strategy for developing a multilevel memory and novel neuromorphic vision systems with integrated sensing-memory-processing.

Photonic Physical Reservoir Computing with Tunable Relaxation Time Constant

Recent years have witnessed a rising demand for edge computing, and there is a need for methods to decrease the computational cost while maintaining a high learning performance when processing information at arbitrary edges. Reservoir computing using physical dynamics has attracted significant attention. However, currently, the timescale of the input signals that can be processed by physical reservoirs is limited by the transient characteristics inherent to the selected physical system. This study used an Sn-doped In2 O3 /Nb-doped SrTiO3 junction to fabricate a memristor that could respond to both electrical and optical stimuli. The results show that the timescale of the transient current response of the device could be controlled over several orders of magnitude simply by applying a small voltage. The computational performance of the device as a physical reservoir is evaluated in an image classification task, demonstrating that the learning accuracy could be optimized by tuning the device to exhibit appropriate transient characteristics according to the timescale of the input signals. These results are expected to provide deeper insights into the photoconductive properties of strontium titanate, as well as support the physical implementation of computing systems.

Filament-free memristors for computing

Memristors have attracted increasing attention due to their tremendous potential to accelerate data-centric computing systems. The dynamic reconfiguration of memristive devices in response to external electrical stimuli can provide highly desirable novel functionalities for computing applications when compared with conventional complementary-metal-oxide-semiconductor (CMOS)-based devices. Those most intensively studied and extensively reviewed memristors in the literature so far have been filamentary type memristors, which typically exhibit a relatively large variability from device to device and from switching cycle to cycle. On the other hand, filament-free switching memristors have shown a better uniformity and attractive dynamical properties, which can enable a variety of new computing paradigms but have rarely been reviewed. In this article, a wide range of filament-free switching memristors and their corresponding computing applications are reviewed. Various junction structures, switching properties, and switching principles of filament-free memristors are surveyed and discussed. Furthermore, we introduce recent advances in different computing schemes and their demonstrations based on non-filamentary memristors. This Review aims to present valuable insights and guidelines regarding the key computational primitives and implementations enabled by these filament-free switching memristors.

Reconfigurable Neuromorphic Computing: Materials, Devices, and Integration

Neuromorphic computing has been attracting ever-increasing attention due to superior energy efficiency, with great promise to promote the next wave of artificial general intelligence in the post-Moore era. Current approaches are, however, broadly designed for stationary and unitary assignments, thus encountering reluctant interconnections, power consumption, and data-intensive computing in that domain. Reconfigurable neuromorphic computing, an on-demand paradigm inspired by the inherent programmability of brain, can maximally reallocate finite resources to perform the proliferation of reproducibly brain-inspired functions, highlighting a disruptive framework for bridging the gap between different primitives. Although relevant research has flourished in diverse materials and devices with novel mechanisms and architectures, a precise overview remains blank and urgently desirable. Herein, the recent strides along this pursuit are systematically reviewed from material, device, and integration perspectives. At the material and device level, one comprehensively conclude the dominant mechanisms for reconfigurability, categorized into ion migration, carrier migration, phase transition, spintronics, and photonics. Integration-level developments for reconfigurable neuromorphic computing are also exhibited. Finally, a perspective on the future challenges for reconfigurable neuromorphic computing is discussed, definitely expanding its horizon for scientific communities.

Reservoir computing with higher-order interactive coupled pendulums

The reservoir computing approach utilizes a time series of measurements as input to a high-dimensional dynamical system known as a reservoir. However, the approach relies on sampling a random matrix to define its underlying reservoir layer, which leads to numerous hyperparameters that need to be optimized. Here, we propose a nonlocally coupled pendulum model with higher-order interactions as a novel reservoir, which requires no random underlying matrices and fewer hyperparameters. We use Bayesian optimization to explore the hyperparameter space within a minimal number of iterations and train the coupled pendulums model to reproduce the chaotic attractors, which simplifies complicated hyperparameter optimization. We illustrate the effectiveness of our technique with the Lorenz system and the Hindmarsh-Rose neuronal model, and we calculate the Pearson correlation coefficients between time series and the Hausdorff metrics in the phase space. We demonstrate the contribution of higher-order interactions by analyzing the interaction between different reservoir configurations and prediction performance, as well as computations of the largest Lyapunov exponents. The chimera state is found as the most effective dynamical regime for prediction. The findings, where we present a new reservoir structure, offer potential applications in the design of high-performance modeling of dynamics in physical systems.

Redox-based ion-gating reservoir consisting of (104) oriented LiCoO2 film, assisted by physical masking

Reservoir computing (RC) is a machine learning framework suitable for processing time series data, and is a computationally inexpensive and fast learning model. A physical reservoir is a hardware implementation of RC using a physical system, which is expected to become the social infrastructure of a data society that needs to process vast amounts of information. Ion-gating reservoirs (IGR) are compact and suitable for integration with various physical reservoirs, but the prediction accuracy and operating speed of redox-IGRs using WO3 as the channel are not sufficient due to irreversible Li+ trapping in the WO3 matrix during operation. Here, in order to enhance the computation performance of redox-IGRs, we developed a redox-based IGR using a (104) oriented LiCoO2 thin film with high electronic and ionic conductivity as a trap-free channel material. The subject IGR utilizes resistance change that is due to a redox reaction (LiCoO2 ⟺ Li1-xCoO2 + xLi+ + xe-) with the insertion and desertion of Li+. The prediction error in the subject IGR was reduced by 72% and the operation speed was increased by 4 times compared to the previously reported WO3, which changes are due to the nonlinear and reversible electrical response of LiCoO2 and the high dimensionality enhanced by a newly developed physical masking technique. This study has demonstrated the possibility of developing high-performance IGRs by utilizing materials with stronger nonlinearity and by increasing output dimensionality.