systems, which can be integrated with terminal sensors to form intelligent sensory systems. [3-7] Information processing in biological sensory nervous systems involves billions of neurons interconnected through trillions of synapses, constituting immense neural networks. [8] Compared with traditional von Neumannbased computing architecture, neural networks greatly reduce time-and energy consumption by taking the advantage of co-location of logic and memory, hyperconnectivity, robustness, and massively parallel processing. [9] As the third-generation artificial neural network, spiking neural networks (SNNs) are inspired by biological nervous systems. They employ spiking neurons as computational units that process information with timing of spikes. [10] Therefore, SNNs provide the potential for spatiotemporal information processing with high time-and energy-efficiency. The learning and recognition functions aiming at spatiotemporal patterns have been demonstrated in SNN formed by resistive switching synaptic devices. [11] However, this synaptic device suffers from several performance limitations such as the discrete weight update, write variation, and high energy programing related to the filamentary switching mechanism. [12-17] Recently, He et al. use capacitively coupled multi-terminal neurotransistors for spatiotemporal information processing, where the neurotransistors mimic the dendritic discriminability of different spatiotemporal input sequences. [18] To physically emulate neural networks in hardware, large-scale crossbar Spiking neural networks (SNNs) sharing large similarity with biological nervous systems are promising to process spatiotemporal information and can provide highly time-and energy-efficient computational paradigms for the Internet-of-Things and edge computing. Nonvolatile electrolyte-gated transistors (EGTs) provide prominent analog switching performance, the most critical feature of synaptic element, and have been recently demonstrated as a promising synaptic device. However, high performance, large-scale EGT arrays, and EGT application for spatiotemporal information processing in an SNN are yet to be demonstrated. Here, an oxide-based EGT employing amorphous Nb 2 O 5 and Li x SiO 2 is introduced as the channel and electrolyte gate materials, respectively, and integrated into a 32 × 32 EGT array. The engineered EGTs show a quasi-linear update, good endurance (10 6) and retention, a high switching speed of 100 ns, ultralow readout conductance (<100 nS), and ultralow areal switching energy density (20 fJ µm −2). The prominent analog switching performance is leveraged for hardware implementation of an SNN with the capability of spatiotemporal information processing, where spike sequences with different timings are able to be efficiently learned and recognized by the EGT array. Finally, this EGT-based spatiotemporal information processing is deployed to detect moving orientation in a tactile sensing system. These results provide an insight into oxide-based EGT devices for energy-efficient neuro...
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 (e.g., visual, pressure, sound, etc.) 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.
Two-dimensional (2D) materials and van der Waals heterostructures have attracted tremendous attention because of their appealing electronic, mechanical, and optoelectronic properties, which offer the possibility to extend the range of functionalities for diverse potential applications. Here, we fabricate a novel multiterminal device with dual-gate based on 2D material van der Waals heterostructures. Such a multiterminal device exhibited excellent nonvolatile multilevel resistance switching performance controlled by the source–drain voltage and back-gate voltage. Based on these features, heterosynaptic plasticity, in which the synaptic weight can be tuned by another modulatory interneuron, has been mimicked. A tunable analogue weight update (both on/off ratio and update nonlinearity) of synapse with high speed (50 ns) and low energy (∼7.3 fJ) programming has been achieved. These results demonstrate the great potential of the artificial synapse based on van der Waals heterostructures for neuromorphic computing.
The sensory nervous system (SNS) builds up the association between external stimuli and the response of organisms. In this system, habituation is a fundamental characteristic that filters out irrelevantly repetitive information and makes the SNS adapt to the external environment. To emulate this critical process in electronic devices, a LixSiOy‐based memristor (TiN/LixSiOy/Pt) is developed where the temporal response under repetitive stimulation is similar to that of habituation. By connecting this synaptic device to a leaky integrate‐and‐fire neuron based on a Ag/SiO2:Ag/Au memristor, a fully memristive SNS with habituation is experimentally demonstrated. Finally, a habituation spiking neural network based on the SNS is built and its application in obstacle avoidance for robot navigation is successfully presented. The results provide that a direct emulation of the biologically inspired learning process by memristors could be a sound choice for neuromorphic hardware implementation.
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 Hf0.2Zr0.8O2 anti-ferroelectric film to meet these challenges. The intrinsic accumulated polarization/spontaneous depolarization of Hf0.2Zr0.8O2 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 (>1012), 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.
Neuromorphic computing powered by spiking neural networks (SNN) provides a powerful and efficient information processing paradigm. To harvest the advantage of SNNs, compact and low‐power synapses that can reliably practice local learning rules are required, posing significant challenges to the conventional silicon‐based platform in terms of area‐ and energy‐efficiency, as well as computing throughput. Here, electrolyte‐gated transistors (EGTs) paired with transistors are employed to implement power‐efficient neuromorphic computing systems. The one‐transistor‐one‐EGT (1T1E) synapse not only alleviates the self‐discharging of EGT but also provides a flexible and efficient way to practice the important spike‐timing‐dependent plasticity learning rule. Based on that, an SNN with a temporal coding scheme is implemented for associative memory that can learn and recover images of handwritten digits with high robustness. Thanks to the temporal coding scheme and low operation current of EGTs, the energy‐efficiency of 1T1E‐based SNN is ≈30× lower than that of the prevalent rate coding scheme, and the peak performance is estimated to be 2 pJ/SOP (picojoule per synaptic operation) at the training phase and 80 TOPs−1 W−1 (tera operations per second per watt) at inference phase, respectively. These results pave the way for power‐efficient neuromorphic computing systems with wide applications for edge computing.
Electrolyte-gated transistors (EGTs) provide prominent analog switching performance for neuromorphic computing. However, suffering from self-discharging nature, the retention performance greatly hampers their practical applications. In this Letter, we realize a significant improvement in EGT retention by inserting a SiO2 layer between the gate electrode and electrolyte. The dynamic process behind the improvement is interpreted by an assumptive leakage-assisted electrochemical mechanism. In addition to improved retention, analog switching with a large dynamic range, superior linearity and symmetry, and low variation has been achieved using identical voltage pulses. Based on the experimental data, a nearly ideal recognition accuracy of 98% has been demonstrated by simulations using the handwritten digit data sets. The obtained results pave a way for employing EGT in future neuromorphic computing.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.