\n Computers are nowhere near as versatile as our own brains. Merolla\n et al.\n applied our present knowledge of the structure and function of the brain to design a new computer chip that uses the same wiring rules and architecture. The flexible, scalable chip operated efficiently in real time, while using very little power.\n

The proper functioning of the adult mammalian brain relies on the orchestrated regulation of neural activity by a diverse population of GABA (gamma-aminobutyric acid)-releasing neurons. Until recently, our appreciation of GABA-mediated inhibition focused predominantly on the GABA(A) (GABA type A) receptors located at synaptic contacts, which are activated in a transient or 'phasic' manner by GABA that is released from synaptic vesicles. However, there is growing evidence that low concentrations of ambient GABA can persistently activate certain subtypes of GABA(A) receptor, which are often remote from synapses, to generate a 'tonic' conductance. In this review, we consider the distinct roles of synaptic and extrasynaptic GABA receptor subtypes in the control of neuronal excitability.

Mammals adapt to a rapidly changing world because of the sophisticated cognitive functions that are supported by the neocortex. The neocortex, which forms almost 80% of the human brain, seems to have arisen from repeated duplication of a stereotypical microcircuit template with subtle specializations for different brain regions and species. The quest to unravel the blueprint of this template started more than a century ago and has revealed an immensely intricate design. The largest obstacle is the daunting variety of inhibitory interneurons that are found in the circuit. This review focuses on the organizing principles that govern the diversity of inhibitory interneurons and their circuits.

We discuss the biological plausibility and computational efficiency of some of the most useful models of spiking and bursting neurons. We compare their applicability to large-scale simulations of cortical neural networks.

Fura-2 calcium imaging in the cricket omega neuron revealed increased intracellular free calcium ion concentration in response to simulated cricket calling songs and other sound stimuli. The time course of the increase and decrease in intracellular calcium coincided with the time course of forward masking, a time-dependent modulation of auditory sensitivity. The buffering of calcium transients with high concentrations of a kinetically fast calcium buffer eliminated the post-stimulus hyperpolarization associated with forward masking, whereas the uncaging of calcium inside the neuron produced a hyperpolarization. The results suggest that sound-stimulated intracellular calcium accumulation acts by means of a calcium-activated hyperpolarizing current to produce forward masking. These findings underscore the importance of chemical dynamics in neural computation by demonstrating a behaviorally relevant role of calcium dynamics in vivo.

How active membrane conductance dynamics tunes neurons for specific time-varying stimuli remains poorly understood. We studied the biophysical mechanisms by which spike frequency adaptation shapes visual stimulus selectivity in an identified visual interneuron of the locust. The lobula giant movement detector (LGMD) responds preferentially to objects approaching on a collision course with the locust. Using calcium imaging, pharmacology and modeling, we show that spike frequency adaptation in the LGMD is mediated by a Ca(2+)-dependent potassium conductance closely resembling those associated with 'small-conductance' (SK) channels. Intracellular block of this conductance minimally affected the LGMD's response to approaching stimuli, but substantially increased its response to translating ones. Thus, spike frequency adaptation contributes to the neuron's tuning by selectively decreasing its responses to nonpreferred stimuli. Our results identify a new mechanism by which spike frequency adaptation may tune visual neurons to behaviorally relevant stimuli.

Cortical processing reflects the interplay of synaptic excitation and synaptic inhibition. Rapidly accumulating evidence is highlighting the crucial role of inhibition in shaping spontaneous and sensory-evoked cortical activity and thus underscores how a better knowledge of inhibitory circuits is necessary for our understanding of cortical function. We discuss current views of how inhibition regulates the function of cortical neurons and point to a number of important open questions.Copyright © 2011 Elsevier Inc. All rights reserved.

Newborn animals, such as 4-month-old infants, 4-day-old chicks, and 1-day-old guppies, exhibit sensitivity to an approximate number of items in the visual array. These findings are often interpreted as evidence for an innate “number sense.” However, number sense is typically investigated using explicit behavioral tasks, which require a form of calibration (e.g., habituation or reward-based training) in experimental studies. Therefore, the generation of number sense may be the result of calibration. We built a number-sense neural network model on the basis of lateral inhibition to explore whether animals demonstrate an innate “number sense” and determine important factors affecting this competence. The proposed model can reproduce size and distance effects of output responses of number-selective neurons when network connection weights are set randomly without an adjustment. Results showed that number sense can be produced under the influence of lateral inhibition, which is one of the fundamental mechanisms of the nervous system, and independent of learning.

In this work, we fabricated a dual gate positive feedback field-effect transistor (FBFET) integrated with CMOS. We investigated the DC and transient characteristics of the FBFET. The fabricated FBFET has an extremely low sub-threshold slope of less than 2.3 mV/dec and low off-current. We also propose an analog integrated-and-fire neuron circuit incorporating a FBFET, which significantly reduces the power dissipation of hardware neural networks. In a conventional neuron circuit using a membrane capacitor to integrate input pulses, most of the energy is consumed by the first inverter stage connected to the capacitor. Since the membrane capacitor is charged slowly compared to digital logic, a large amount of short-circuit current flows between Vdd and ground in the first inverter during this period. In the proposed neuron circuit, the short-circuit current is significantly suppressed by adopting a FBFET in the inverter. Through TCAD mixed mode simulation of the device and the circuit, we compare the energy consumption of a conventional and the proposed neuron circuits. In a single neuron circuit with microsecond duration pulses, 58% of the energy consumption is reduced by incorporating a FBFET. We performed SPICE compact modeling of FBFET, and its parameters were fitted to match the measurement results of the fabricated FBFET. Then, we conducted a circuit simulation to verify the operating neural networks. We implemented a single layer spiking neural network (SNN) that had resistive synaptic devices. In the SNN simulation, approximately 94% of the average power consumption of all output neurons was reduced.

The Hodgkin-Huxley model for action potential generation in biological axons is central for understanding the computational capability of the nervous system and emulating its functionality. Owing to the historical success of silicon complementary metal-oxide-semiconductors, spike-based computing is primarily confined to software simulations and specialized analogue metal-oxide-semiconductor field-effect transistor circuits. However, there is interest in constructing physical systems that emulate biological functionality more directly, with the goal of improving efficiency and scale. The neuristor was proposed as an electronic device with properties similar to the Hodgkin-Huxley axon, but previous implementations were not scalable. Here we demonstrate a neuristor built using two nanoscale Mott memristors, dynamical devices that exhibit transient memory and negative differential resistance arising from an insulating-to-conducting phase transition driven by Joule heating. This neuristor exhibits the important neural functions of all-or-nothing spiking with signal gain and diverse periodic spiking, using materials and structures that are amenable to extremely high-density integration with or without silicon transistors.

Neuromorphic networks of artificial neurons and synapses can solve computationally hard problems with energy efficiencies unattainable for von Neumann architectures. For image processing, silicon neuromorphic processors outperform graphic processing units in energy efficiency by a large margin, but deliver much lower chip-scale throughput. The performance-efficiency dilemma for silicon processors may not be overcome by Moore’s law scaling of silicon transistors. Scalable and biomimetic active memristor neurons and passive memristor synapses form a self-sufficient basis for a transistorless neural network. However, previous demonstrations of memristor neurons only showed simple integrate-and-fire behaviors and did not reveal the rich dynamics and computational complexity of biological neurons. Here we report that neurons built with nanoscale vanadium dioxide active memristors possess all three classes of excitability and most of the known biological neuronal dynamics, and are intrinsically stochastic. With the favorable size and power scaling, there is a path toward an all-memristor neuromorphic cortical computer.

Pb(Zr0.53Ti0.47)O3 (PZT) thin film was fabricated on Pt/Ti/SiO2/Si substrate by chemical solution deposition method. Our results show a very great switchable ferroelectric diode effect (SFDE) in Pt-PZT-Au structure, which is more obvious and controllable than that in other ferroelectric thin films. The electrical conduction exhibits high rectifying behavior after pre-poling and the polarity of ferroelectric diode can be switched by changing the orientation of polarization in ferroelectric thin film. Our results also indicate that the SFDE in PZT film is highly dependent on remanent polarization and temperature. With the increase of remanent polarization, the forward current of bistable rectifying behavior observably reduces. Therefore, our measurement indicated that the biggest rectification ratio can reach about 220, which is found in 250K after +10V poling. By analyzing the conduction data, it is found that the dominant conduction mechanism of the SFDE in this sample is due to the space-charge-limited bulk conduction (SCLC), and Schottky emission (SE) may play subordinate role in forward bias voltage. Our observation demonstrates that SFDE may be general characteristic in ferroelectrics as long as proper electrodes chosen.

BiFeO 3 was studied as an alternative environmentally clean ferro/piezoelectric material. 200-nm-thick BiFeO3 films were grown on Si substrates with SrTiO3 as a template layer and SrRuO3 as bottom electrode. X-ray and transmission electron microscopy studies confirmed the epitaxial growth of the films. The spontaneous polarization of the films was ∼45μC∕cm2. Retention measurement up to several days showed no decay of polarization. A piezoelectric coefficient (d33) of ∼60pm∕V was observed, which is promising for applications in micro-electro-mechanical systems and actuators.

We report that crystalline phases with ferroelectric behavior can be formed in thin films of SiO2 doped hafnium oxide. Films with a thickness of 10 nm and with less than 4 mol. % of SiO2 crystallize in a monoclinic/tetragonal phase mixture. We observed that the formation of the monoclinic phase is inhibited if crystallization occurs under mechanical encapsulation and an orthorhombic phase is obtained. This phase shows a distinct piezoelectric response, while polarization measurements exhibit a remanent polarization above 10 μC/cm2 at a coercive field of 1 MV/cm, suggesting that this phase is ferroelectric. Ferroelectric hafnium oxide is ideally suited for ferroelectric field effect transistors and capacitors due to its excellent compatibility to silicon technology.

Neuron is the basic computing unit in brain-inspired neural networks. Although a multitude of excellent artificial neurons realized with conventional transistors have been proposed, they might not be energy and area efficient in large-scale networks. The recent discovery of ferroelectricity in hafnium oxide (HfO2) and the related switching phenomena at the nanoscale might provide a solution. This study employs the newly reported accumulative polarization reversal in nanoscale HfO2-based ferroelectric field-effect transistors (FeFETs) to implement two key neuronal dynamics: the integration of action potentials and the subsequent firing according to the biologically plausible all-or-nothing law. We show that by carefully shaping electrical excitations based on the particular nucleation-limited switching kinetics of the ferroelectric layer further neuronal behaviors can be emulated, such as firing activity tuning, arbitrary refractory period and the leaky effect. Finally, we discuss the advantages of an FeFET-based neuron, highlighting its transferability to advanced scaling technologies and the beneficial impact it may have in reducing the complexity of neuromorphic circuits.

Neuromorphic machines are intriguing for building energy-efficient intelligent systems, where spiking neurons are pivotal components. Recently, memristive neurons with promising bio-plausibility have been developed, but with limited reliability, bulky capacitors or additional reset circuits. Here, we propose an anti-ferroelectric field-effect transistor neuron based on the inherent polarization and depolarization of HfZrO anti-ferroelectric film to meet these challenges. The intrinsic accumulated polarization/spontaneous depolarization of HfZrO films implements the integration/leaky behavior of neurons, avoiding external capacitors and reset circuits. Moreover, the anti-ferroelectric neuron exhibits low energy consumption (37 fJ/spike), high endurance (>10), high uniformity and high stability. We further construct a two-layer fully ferroelectric spiking neural networks that combines anti-ferroelectric neurons and ferroelectric synapses, achieving 96.8% recognition accuracy on the Modified National Institute of Standards and Technology dataset. This work opens the way to emulate neurons with anti-ferroelectric materials and provides a promising approach to building high-efficient neuromorphic hardware.© 2022. The Author(s).

One long-standing goal in the emerging neuromorphic field is to create a reliable neural network hardware implementation that has low energy consumption, while providing massively parallel computation. Although diverse oxide-based devices have made significant progress as artificial synaptic and neuronal components, these devices still need further optimization regarding linearity, symmetry, and stability. Here, we present a proof-of-concept experiment for integrated neuromorphic computing networks by utilizing spintronics-based synapse (spin-S) and neuron (spin-N) devices, along with linear and symmetric weight responses for spin-S using a stripe domain and activation functions for spin-N. An integrated neural network of electrically connected spin-S and spin-N successfully proves the integration function for a simple pattern classification task. We simulate a spin-N network using the extracted device characteristics and demonstrate a high classification accuracy (over 93%) for the spin-S and spin-N optimization without the assistance of additional software or circuits required in previous reports. These experimental studies provide a new path toward establishing more compact and efficient neural network systems with optimized multifunctional spintronic devices.

Neuromorphic computing using nonvolatile memories is expected to tackle the memory wall and energy efficiency bottleneck in the von Neumann system and to mitigate the stagnation of Moore's law. However, an ideal artificial neuron possessing bio-inspired behaviors as exemplified by the requisite leaky-integrate-fire and self-reset (LIFT) functionalities within a single device is still lacking. Here, we report a new type of spiking neuron with LIFT characteristics by manipulating the magnetic domain wall motion in a synthetic antiferromagnetic (SAF) heterostructure. We validate the mechanism of Joule heating modulated competition between the Ruderman-Kittel-Kasuya-Yosida interaction and the built-in field in the SAF device, enabling it with a firing rate up to 17 MHz and energy consumption of 486 fJ/spike. A spiking neuron circuit is implemented with a latency of 170 ps and power consumption of 90.99 μW. Moreover, the winner-takes-all is executed with a current ratio >10 between activated and inhibited neurons. We further establish a two-layer spiking neural network based on the developed spintronic LIFT neurons. The architecture achieves 88.5% accuracy on the handwritten digit database benchmark. Our studies corroborate the circuit compatibility of the spintronic neurons and their great potential in the field of intelligent devices and neuromorphic computing.© 2023. The Author(s).

Nanoscale multilayer structure TiO2/BaTiO3/TiO2 has been fabricated on Pt/Ti/SiO2/Si substrate by chemical solution deposition method. Highly uniform bipolar resistive switching (BRS) characteristics have been observed in Pt/TiO2/BaTiO3/TiO2/Pt cells. Analysis of the current-voltage relationship demonstrates that the space-charge-limited current conduction controlled by the localized oxygen vacancies should be important to the resistive switching behavior. X-ray photoelectron spectroscopy results indicated that oxygen vacancies in TiO2 play a crucial role in the resistive switching phenomenon and the introduced TiO2/BaTiO3 interfaces result in the high uniformity of bipolar resistive switching characteristics.

It has been reported that in chalcogenide-based electrochemical metallization (ECM) memory cells (e.g., As2S3:Ag, GeS:Cu, and Ag2S), the metal filament grows from the cathode (e.g., Pt and W) towards the anode (e.g., Cu and Ag), whereas filament growth along the opposite direction has been observed in oxide-based ECM cells (e.g., ZnO, ZrO2, and SiO2). The growth direction difference has been ascribed to a high ion diffusion coefficient in chalcogenides in comparison with oxides. In this paper, upon analysis of OFF state I–V characteristics of ZnS-based ECM cells, we find that the metal filament grows from the anode towards the cathode and the filament rupture and rejuvenation occur at the cathodic interface, similar to the case of oxide-based ECM cells. It is inferred that in ECM cells based on the chalcogenides such as As2S3:Ag, GeS:Cu, and Ag2S, the filament growth from the cathode towards the anode is due to the existence of an abundance of ready-made mobile metal ions in the chalcogenides rather than to the high ion diffusion coefficient.

Memristor, fusing the functions of storage and computing within a single device, is one of the core electronic components to solve the bottleneck of von Neumann architecture. With the unique volatile/non-volatile resistive switching characteristic, memristor can simulate the function of synapses/neurons in brain well. In addition, due to the compatibility with traditional complementary metal-oxide-semiconductor (CMOS) processes, metal-oxide-based memristors have received a lot of attention. In recent years, many kinds of metal-oxide memristors based on single dielectric layer have been proposed. However, there are still some problems such as the instability of switching voltage, fluctuation of high/low resistance state and poor endurance of memristive device. Thus, the researchers have successfully optimized the device performance by introducing the double dielectric layer into the metal-oxide memristors. In this article, we introduce the advantages of double dielectric layers-based metal-oxide memristors, and discuss their mechanism and design of double dielectric layers-based metal-oxide memristors. Eventually, we introduce their potential applications in neuromorphic computing. This review provides some enlightenment on how to design high-performance metal-oxide memristor based on double dielectric layers.

Although there is a huge progress in complementary-metal-oxide-semiconductor (CMOS) technology, construction of an artificial neural network using CMOS technology to realize the functionality comparable with that of human cerebral cortex containing 1010–1011 neurons is still of great challenge. Recently, phase change memristor neuron has been proposed to realize a human-brain level neural network operating at a high speed while consuming a small amount of power and having a high integration density. Although memristor neuron can be scaled down to nanometer, integration of 1010–1011 neurons still faces many problems in circuit complexity, chip area, power consumption, etc. In this work, we propose a CMOS compatible HfO2 memristor neuron that can be well integrated with silicon circuits. A hybrid Convolutional Neural Network (CNN) based on the HfO2 memristor neuron is proposed and constructed. In the hybrid CNN, one memristive neuron can behave as multiple physical neurons based on the Time Division Multiplexing Access (TDMA) technique. Handwritten digit recognition is demonstrated in the hybrid CNN with a memristive neuron acting as 784 physical neurons. This work paves the way towards substantially shrinking the amount of neurons required in hardware and realization of more complex or even human cerebral cortex level memristive neural networks.

Artificial neuromorphic systems based on populations of spiking neurons are an indispensable tool in understanding the human brain and in constructing neuromimetic computational systems. To reach areal and power efficiencies comparable to those seen in biological systems, electroionics-based and phase-change-based memristive devices have been explored as nanoscale counterparts of synapses. However, progress on scalable realizations of neurons has so far been limited. Here, we show that chalcogenide-based phase-change materials can be used to create an artificial neuron in which the membrane potential is represented by the phase configuration of the nanoscale phase-change device. By exploiting the physics of reversible amorphous-to-crystal phase transitions, we show that the temporal integration of postsynaptic potentials can be achieved on a nanosecond timescale. Moreover, we show that this is inherently stochastic because of the melt-quench-induced reconfiguration of the atomic structure occurring when the neuron is reset. We demonstrate the use of these phase-change neurons, and their populations, in the detection of temporal correlations in parallel data streams and in sub-Nyquist representation of high-bandwidth signals.

A major challenge in the field of neurocomputing is to mimic the brain's behavior by implementing artificial synapses and neurons directly in hardware. Toward that purpose, many researchers are exploring the potential of new materials and new physical phenomena. Recently, a new concept of the Leaky Integrate and Fire (LIF) artificial neuron was proposed based on the electric Mott transition in the inorganic Mott insulator GaTa4Se8. In this work, we report on the LIF behavior in simple two-terminal devices in three chemically very different compounds, the oxide (V0.89Cr0.11)2O3, the sulfide GaMo4S8, and the molecular system [Au(iPr-thiazdt)2] (C12H14AuN2S8), but sharing a common feature, their Mott insulator ground state. In all these devices, the application of an electric field induces a volatile resistive switching and a remarkable LIF behavior under a train of pulses. It suggests that the Mott LIF neuron is a general concept that can be extended to the large class of Mott insulators.

Artificial neural networks can harness stochasticity in multiple ways to enable a vast class of computationally powerful models. Boltzmann machines and other stochastic neural networks have been shown to outperform their deterministic counterparts by allowing dynamical systems to escape local energy minima. Electronic implementation of such stochastic networks is currently limited to addition of algorithmic noise to digital machines which is inherently inefficient; albeit recent efforts to harness physical noise in devices for stochasticity have shown promise. To succeed in fabricating electronic neuromorphic networks we need experimental evidence of devices with measurable and controllable stochasticity which is complemented with the development of reliable statistical models of such observed stochasticity. Current research literature has sparse evidence of the former and a complete lack of the latter. This motivates the current article where we demonstrate a stochastic neuron using an insulator-metal-transition (IMT) device, based on electrically induced phase-transition, in series with a tunable resistance. We show that an IMT neuron has dynamics similar to a piecewise linear FitzHugh-Nagumo (FHN) neuron and incorporates all characteristics of a spiking neuron in the device phenomena. We experimentally demonstrate spontaneous stochastic spiking along with electrically controllable firing probabilities using Vanadium Dioxide (VO) based IMT neurons which show a sigmoid-like transfer function. The stochastic spiking is explained by two noise sources - thermal noise and threshold fluctuations, which act as precursors of bifurcation. As such, the IMT neuron is modeled as an Ornstein-Uhlenbeck (OU) process with a fluctuating boundary resulting in transfer curves that closely match experiments. The moments of interspike intervals are calculated analytically by extending the first-passage-time (FPT) models for Ornstein-Uhlenbeck (OU) process to include a fluctuating boundary. We find that the coefficient of variation of interspike intervals depend on the relative proportion of thermal and threshold noise, where threshold noise is the dominant source in the current experimental demonstrations. As one of the first comprehensive studies of a stochastic neuron hardware and its statistical properties, this article would enable efficient implementation of a large class of neuro-mimetic networks and algorithms.

The accumulation and extrusion of Ca in the pre- and postsynaptic compartments play a critical role in initiating plastic changes in biological synapses. To emulate this fundamental process in electronic devices, we developed diffusive Ag-in-oxide memristors with a temporal response during and after stimulation similar to that of the synaptic Ca dynamics. In situ high-resolution transmission electron microscopy and nanoparticle dynamics simulations both demonstrate that Ag atoms disperse under electrical bias and regroup spontaneously under zero bias because of interfacial energy minimization, closely resembling synaptic influx and extrusion of Ca, respectively. The diffusive memristor and its dynamics enable a direct emulation of both short- and long-term plasticity of biological synapses, representing an advance in hardware implementation of neuromorphic functionalities.

The cone photoreceptors in our eyes selectively transduce the natural light into spiking representations, which endows the brain with high energy-efficiency color vision. However, the cone-like device with color-selectivity and spike-encoding capability remains challenging. Here, we propose a metal oxide-based vertically integrated spiking cone photoreceptor array, which can directly transduce persistent lights into spike trains at a certain rate according to the input wavelengths. Such spiking cone photoreceptors have an ultralow power consumption of less than 400 picowatts per spike in visible light, which is very close to biological cones. In this work, lights with three wavelengths were exploited as pseudo-three-primary colors to form 'colorful' images for recognition tasks, and the device with the ability to discriminate mixed colors shows better accuracy. Our results would enable hardware spiking neural networks with biologically plausible visual perception and provide great potential for the development of dynamic vision sensors.© 2023. The Author(s).

Two methods for classification based on the Bayes strategy and nonparametric estimators for probability density functions are reviewed. The two methods are named the probabilistic neural network (PNN) and the polynomial Adaline. Both methods involve one-pass learning algorithms that can be implemented directly in parallel neural network architectures. The performances of the two methods are compared with multipass backpropagation networks, and relative advantages and disadvantages are discussed. PNN and the polynomial Adaline are complementary techniques because they implement the same decision boundaries but have different advantages for applications. PNN is easy to use and is extremely fast for moderate-sized databases. For very large databases and for mature applications in which classification speed is more important than training speed, the polynomial equivalent can be found.

In recent decades, artificial intelligence has been successively employed in the fields of finance, commerce, and other industries. However, imitating high-level brain functions, such as imagination and inference, pose several challenges as they are relevant to a particular type of noise in a biological neuron network. Probabilistic computing algorithms based on restricted Boltzmann machine and Bayesian inference that use silicon electronics have progressed significantly in terms of mimicking probabilistic inference. However, the quasi-random noise generated from additional circuits or algorithms presents a major challenge for silicon electronics to realize the true stochasticity of biological neuron systems. Artificial neurons based on emerging devices, such as memristors and ferroelectric field-effect transistors with inherent stochasticity can produce uncertain non-linear output spikes, which may be the key to make machine learning closer to the human brain. In this article, we present a comprehensive review of the recent advances in the emerging stochastic artificial neurons (SANs) in terms of probabilistic computing. We briefly introduce the biological neurons, neuron models, and silicon neurons before presenting the detailed working mechanisms of various SANs. Finally, the merits and demerits of silicon-based and emerging neurons are discussed, and the outlook for SANs is presented.

Biological neurons exhibit dynamic excitation behavior in the form of stochastic firing, rather than stiffly giving out spikes upon reaching a fixed threshold voltage, which empowers the brain to perform probabilistic inference in the face of uncertainty. However, owing to the complexity of the stochastic firing process in biological neurons, the challenge of fabricating and applying stochastic neurons with bio-realistic dynamics to probabilistic scenarios remains to be fully addressed. In this work, a novel CuS/GeSe conductive-bridge threshold switching memristor is fabricated and singled out to realize electronic stochastic neurons, which is ascribed to the similarity between the stochastic switching behavior observed in the device and that of biological ion channels. The corresponding electric circuit of a stochastic neuron is then constructed and the probabilistic firing capacity of the neuron is utilized to implement Bayesian inference in a spiking neural network (SNN). The application prospects are demonstrated on the example of a tumor diagnosis task, where common fatal diagnostic errors of a conventional artificial neural network are successfully circumvented. Moreover, in comparison to deterministic neuron-based SNNs, the stochastic neurons enable SNNs to deliver an estimate of the uncertainty in their predictions, and the fidelity of the judgement is drastically improved by 81.2%.

Neuromorphic architectures based on memristive neurons and synapses hold great prospect in achieving highly intelligent and efficient computing systems. Here, we show that a Schottky diode based on Cu-Ta/InGaZnO4 (IGZO)/TiN structure can exhibit threshold switching behavior after electroforming and in turn be used to implement an artificial neuron with inherently stochastic dynamics. The threshold switching originates from the Cu filament formation and spontaneous Cu–In–O precipitation in IGZO. The nucleation and precipitation of Cu–In–O phase are stochastic in nature, which leads to the stochasticity of the artificial neuron. It is demonstrated that IGZO based stochastic neurons can be used for global minimum computation with random walk algorithm, making it promising for robust neuromorphic computation.

Threshold switching (TS) devices are finding increasing use in the hardware implementation of neuromorphic network computing. Here, a simple structured Ag/amorphous Si/Pt TS device with a switching ratio of ∼105 is prepared, with turn-on and turn-off speeds as high as ∼20 ns and ∼16 ns, respectively. We use this TS device to construct a leaky integration-and-firing artificial neuron that emulates key biological neuron features like threshold-driven firing, all-or-nothing spiking, refractory period, intensity-modulated frequency response, and conductance-modulated frequency response. These results suggest that Si film-based TS device artificial neurons have significant potential for building high-speed artificial neural networks.

A spiking neural network consists of artificial synapses and neurons and may realize human-level intelligence. Unlike the widely reported artificial synapses, the fabrication of large-scale artificial neurons with good performance is still challenging due to the lack of a suitable material system and integration method. Here, we report an ultrathin (less than10 nm) and inch-size two-dimensional (2D) oxide-based artificial neuron system produced by a controllable assembly of solution-processed 2D monolayer TiO nanosheets. Artificial neuron devices based on such 2D TiO films show a high on/off ratio of 10 and a volatile resistance switching phenomenon. The devices can not only emulate the leaky integrate-and-fire activity but also self-recover without additional circuits for sensing and reset. Moreover, the artificial neuron arrays are fabricated and exhibited good uniformity, indicating their large-area integration potential. Our results offer a strategy for fabricating large-scale and ultrathin 2D material-based artificial neurons and 2D spiking neural networks.

The exploration of flexible electronic devices with information processing functions of biological neurons is of great significance for the development of intelligent wearable technologies. Due to lack of inherent mechanical flexibility, conventional threshold-switching memristor based on rigid materials that can implement the computing functions of biological neurons is difficult to fulfill the requirements for potential applications in the future. In this work, an intrinsically stretchable threshold-switching memristor was prepared by using silver nanowire-polyurethane composite as the dielectric layer and liquid metal as the electrodes, respectively. Under application of a sweeping voltage, the device exhibited reliable threshold switching characteristics, which was switched from the high resistance state (HRS) to the low resistance state (LRS) during device programming and spontaneously relaxed to the HRS upon voltage application. Further analysis shows that the underlying mechanism can be attributed to the dynamic formation and rupture of discontinuous silver conductive filaments formed between silver nanowires. In the pulse programming mode, memristor device is able to emulate the integration and firing characteristics of biological neurons, suggesting its great potential as an artificial neuron. Moreover, the pulse amplitude and pulse interval modulated neuronal spiking behaviors are successfully replicated using such devices. Under 20% tensile strain, the threshold-switching memristor shows negligible changes in the operating parameters during device switching and neuronal function implementations, suggesting its excellent mechanical flexibility and stability. This work provides important guidelines for the development of high-performance stretchable artificial neuronal devices and next-generation intelligent wearable systems.

Drawing inspiration from biology, neuromorphic systems are of great interest in direct interaction and efficient processing of analogue signals in the real world and could be promising for the development of smart sensors. Here, we demonstrate an artificial sensory neuron consisting of an InGaZnO (IGZO)-based optical sensor and NbO-based oscillation neuron in series, which can simultaneously sense the optical information even beyond the visible light region and encode them into electrical impulses. Such artificial vision sensory neurons can convey visual information in a parallel manner analogous to biological vision systems, and the output spikes can be effectively processed by a pulse coupled neural network, demonstrating the capability of image segmentation out of a complex background. This study could facilitate the construction of artificial visual systems and pave the way for the development of light-driven neurorobotics, bioinspired optoelectronics, and neuromorphic computing.

The visual perception system is the most important system for human learning since it receives over 80% of the learning information from the outside world. With the exponential growth of artificial intelligence technology, there is a pressing need for high-energy and area-efficiency visual perception systems capable of processing efficiently the received natural information. Currently, memristors with their elaborate dynamics, excellent scalability, and information ( visual, pressure, sound, ) perception ability exhibit tremendous potential for the application of visual perception. Here, we propose a fully memristor-based artificial visual perception nervous system (AVPNS) which consists of a quantum-dot-based photoelectric memristor and a nanosheet-based threshold-switching (TS) memristor. We use a photoelectric and a TS memristor to implement the synapse and leaky integrate-and-fire (LIF) neuron functions, respectively. With the proposed AVPNS we successfully demonstrate the biological image perception, integration and fire, as well as the biosensitization process. Furthermore, the self-regulation process of a speed meeting control system in driverless automobiles can be accurately and conceptually emulated by this system. Our work shows that the functions of the biological visual nervous system may be systematically emulated by a memristor-based hardware system, thus expanding the spectrum of memristor applications in artificial intelligence.

In a neuromorphic computing system, the complex CMOS neuron circuits have been the bottleneck for efficient implementation of weighted sum operation. The phenomenon of metal-insulator-transition (MIT) in strongly correlated oxides, such as NbO2, has shown the oscillation behavior in recent experiments. In this work, we propose using a MIT device to function as a compact oscillation neuron, achieving the same functionality as the CMOS neuron but occupying a much smaller area. Pt/NbOx/Pt devices are fabricated, exhibiting the threshold switching I-V hysteresis. When the NbOx device is connected with an external resistor (i.e., the synapse), the neuron membrane voltage starts a self-oscillation. We experimentally demonstrate that the oscillation frequency is proportional to the conductance of the synapse, showing its feasibility for integrating the weighted sum current. The switching speed measurement indicates that the oscillation frequency could achieve >33 MHz if parasitic capacitance can be eliminated.

In the brain, the membrane potential of many neurons oscillates in a subthreshold damped fashion and fire when excited by an input frequency that nearly equals their eigen frequency. In this work, we investigate theoretically the artificial implementation of such “resonate-and-fire” neurons by utilizing the magnetization dynamics of a fixed magnetic skyrmion in the free layer of a magnetic tunnel junction (MTJ). To realize firing of this nanomagnetic implementation of an artificial neuron, we propose to employ voltage control of magnetic anisotropy or voltage generated strain as an input (spike or sinusoidal) signal, which modulates the perpendicular magnetic anisotropy. This results in continual expansion and shrinking (i.e., breathing) of a skyrmion core that mimics the subthreshold oscillation. Any subsequent input pulse having an interval close to the breathing period or a sinusoidal input close to the eigen frequency drives the magnetization dynamics of the fixed skyrmion in a resonant manner. The time varying electrical resistance of the MTJ layer due to this resonant oscillation of the skyrmion core is used to drive a Complementary Metal Oxide Semiconductor buffer circuit, which produces spike outputs. By rigorous micromagnetic simulation, we investigate the interspike timing dependence and response to different excitatory and inhibitory incoming input pulses. Finally, we show that such resonate and fire neurons have potential application in coupled nanomagnetic oscillator based associative memory arrays.

Magnetic skyrmions are particle-like topological spin configurations, which can carry binary information and thus are promising building blocks for future spintronic devices. In this work, we investigate the relationship between the skyrmion dynamics and the characteristics of injected current in a skyrmion-based spin torque nano-oscillator, where the excitation source is introduced from a point nano-contact at the center of the nanodisk. It is found that the skyrmion will move away from the center of the nanodisk if it is driven by a spin-polarized current; however, it will return to the initial position in the absence of stimulus. Therefore, we propose a skyrmion-based artificial spiking neuron, which can effectively implement the leaky-integrate-fire operation. We study the feasibility of the skyrmion-based spiking neuron by using micromagnetic simulations. Our results may provide useful guidelines for building future magnetic neural networks with ultra-high density and ultra-low energy consumption.

We report a halide perovskite based flexible threshold-switched memristor with ultra-high speed as an artificial neuron that exhibits excellent leaky integrate-and-fire dynamics and strength-modulated spike frequency response performance.

The brain naturally binds events from different sources in unique concepts. It is hypothesized that this process occurs through the transient mutual synchronization of neurons located in different regions of the brain when the stimulus is presented. This mechanism of ‘binding through synchronization’ can be directly implemented in neural networks composed of coupled oscillators. To do so, the oscillators must be able to mutually synchronize for the range of inputs corresponding to a single class, and otherwise remain desynchronized. Here we show that the outstanding ability of spintronic nano-oscillators to mutually synchronize and the possibility to precisely control the occurrence of mutual synchronization by tuning the oscillator frequencies over wide ranges allows pattern recognition. We demonstrate experimentally on a simple task that three spintronic nano-oscillators can bind consecutive events and thus recognize and distinguish temporal sequences. This work is a step forward in the construction of neural networks that exploit the non-linear dynamic properties of their components to perform brain-inspired computations.

Neuromorphic computing based on spikes offers great potential in highly efficient computing paradigms. Recently, several hardware implementations of spiking neural networks based on traditional complementary metal-oxide semiconductor technology or memristors have been developed. However, an interface (called an afferent nerve in biology) with the environment, which converts the analog signal from sensors into spikes in spiking neural networks, is yet to be demonstrated. Here we propose and experimentally demonstrate an artificial spiking afferent nerve based on highly reliable NbO Mott memristors for the first time. The spiking frequency of the afferent nerve is proportional to the stimuli intensity before encountering noxiously high stimuli, and then starts to reduce the spiking frequency at an inflection point. Using this afferent nerve, we further build a power-free spiking mechanoreceptor system with a passive piezoelectric device as the tactile sensor. The experimental results indicate that our afferent nerve is promising for constructing self-aware neurorobotics in the future.

Neuromorphic perception systems inspired by biology have tremendous potential in efficiently processing multi-sensory signals from the physical world, but a highly efficient hardware element capable of sensing and encoding multiple physical signals is still lacking. Here, we report a spike-based neuromorphic perception system consisting of calibratable artificial sensory neurons based on epitaxial VO2, where the high crystalline quality of VO2 leads to significantly improved cycle-to-cycle uniformity. A calibration resistor is introduced to optimize device-to-device consistency, and to adapt the VO2 neuron to different sensors with varied resistance level, a scaling resistor is further incorporated, demonstrating cross-sensory neuromorphic perception component that can encode illuminance, temperature, pressure and curvature signals into spikes. These components are utilized to monitor the curvatures of fingers, thereby achieving hand gesture classification. This study addresses the fundamental cycle-to-cycle and device-to-device variation issues of sensory neurons, therefore promoting the construction of neuromorphic perception systems for e-skin and neurorobotics.

Oscillatory Neural Networks (ONNs) are currently arousing interest in the research community for their potential to implement very fast, ultra-low-power computing tasks by exploiting specific emerging technologies. From the architectural point of view, ONNs are based on the synchronization of oscillatory neurons in cognitive processing, as occurs in the human brain. As emerging technologies, VO2 and memristive devices show promising potential for the efficient implementation of ONNs. Abundant literature is now becoming available pertaining to the study and building of ONNs based on VO2 devices and resistive coupling, such as memristors. One drawback of direct resistive coupling is that physical resistances cannot be negative, but from the architectural and computational perspective this would be a powerful advantage when interconnecting weights in ONNs. Here we solve the problem by proposing a hardware implementation technique based on differential oscillatory neurons for ONNs (DONNs) with VO2-based oscillators and memristor-bridge circuits. Each differential oscillatory neuron is made of a pair of VO2 oscillators operating in anti-phase. This way, the neurons provide a pair of differential output signals in opposite phase. The memristor-bridge circuit is used as an adjustable coupling function that is compatible with differential structures and capable of providing both positive and negative weights. By combining differential oscillatory neurons and memristor-bridge circuits, we propose the hardware implementation of a fully connected differential ONN (DONN) and use it as an associative memory. The standard Hebbian rule is used for training, and the weights are then mapped to the memristor-bridge circuit through a proposed mapping rule. The paper also introduces some functional and hardware specifications to evaluate the design. Evaluation is performed by circuit-level electrical simulations and shows that the retrieval accuracy of the proposed design is comparable to that of classic Hopfield Neural Networks.

As a key building block of biological cortex, neurons are powerful information processing units and can achieve highly complex nonlinear computations even in individual cells. Hardware implementation of artificial neurons with similar capability is of great significance for the construction of intelligent, neuromorphic systems. Here, we demonstrate an artificial neuron based on NbO volatile memristor that not only realizes traditional all-or-nothing, threshold-driven spiking and spatiotemporal integration, but also enables dynamic logic including XOR function that is not linearly separable and multiplicative gain modulation among different dendritic inputs, therefore surpassing neuronal functions described by a simple point neuron model. A monolithically integrated 4 × 4 fully memristive neural network consisting of volatile NbO memristor based neurons and nonvolatile TaO memristor based synapses in a single crossbar array is experimentally demonstrated, showing capability in pattern recognition through online learning using a simplified δ-rule and coincidence detection, which paves the way for bio-inspired intelligent systems.

In the nervous system, dendrites, branches of neurons that transmit signals between synapses and soma, play a critical role in processing functions, such as nonlinear integration of postsynaptic signals. The lack of these critical functions in artificial neural networks compromises their performance, for example in terms of flexibility, energy efficiency and the ability to handle complex tasks. Here, by developing artificial dendrites, we experimentally demonstrate a complete neural network fully integrated with synapses, dendrites and soma, implemented using scalable memristor devices. We perform a digit recognition task and simulate a multilayer network using experimentally derived device characteristics. The power consumption is more than three orders of magnitude lower than that of a central processing unit and 70 times lower than that of a typical application-specific integrated circuit chip. This network, equipped with functional dendrites, shows the potential of substantial overall performance improvement, for example by extracting critical information from a noisy background with significantly reduced power consumption and enhanced accuracy.

Coupled oscillators are highly complex dynamical systems, and it is an intriguing concept to use this oscillator dynamics for computation. The idea is not new, but is currently the subject to intense research as part of the quest for “beyond Moore” electronic devices. To a large extent, these efforts are motivated by biological observations: neural systems and mammalian brains, which seem to operate on oscillatory signals. In this paper, we give a survey of oscillator-based computing, with the goal of understanding its promise and limitation for next-generation computing. Our focus will be on the physics of (mostly nanoscale) oscillatory systems and on their characteristics that may enable effective computing.

Nano-oscillators based on phase-transition materials are being explored for the implementation of different non-conventional computing paradigms. In particular, vanadium dioxide (VO2) devices are used to design autonomous non-linear oscillators from which oscillatory neural networks (ONNs) can be developed. In this work, we propose a new architecture for ONNs in which sub-harmonic injection locking (SHIL) is exploited to ensure that the phase information encoded in each neuron can only take two values. In this sense, the implementation of ONNs from neurons that inherently encode information with two-phase values has advantages in terms of robustness and tolerance to variability present in VO2 devices. Unlike conventional interconnection schemes, in which the sign of the weights is coded in the value of the resistances, in our proposal the negative (positive) weights are coded using static inverting (non-inverting) logic at the output of the oscillator. The operation of the proposed architecture is shown for pattern recognition applications.

We derive experimentally based estimates of the energy used by neural mechanisms to code known quantities of information. Biophysical measurements from cells in the blowfly retina yield estimates of the ATP required to generate graded (analog) electrical signals that transmit known amounts of information. Energy consumption is several orders of magnitude greater than the thermodynamic minimum. It costs 10(4) ATP molecules to transmit a bit at a chemical synapse, and 10(6)-10(7) ATP for graded signals in an interneuron or a photoreceptor, or for spike coding. Therefore, in noise-limited signaling systems, a weak pathway of low capacity transmits information more economically, which promotes the distribution of information among multiple pathways.

The cost of not thinking: unused synapses are metabolically expensive due to persistent activity of a synaptic vesicle proton pump.