All-optical Leaky-integrate-and-fire Neuron with Half-sigmoid Transfer Function

Qiang Li,Zhi Wang,Ziling Fu,Huiying Wang,Can Cui,Chongqing Wu

2019-01-01

Abstract:We experimentally demonstrated an all-optical leaky-integrate-and-fire (LIF) neuron which can perform half-sigmoid transfer function. The proposed all-optical LIF neuron can behave as an excitatory or inhibitory neuron by tuning the initial state of polarization.

What problem does this paper attempt to address?

Artificial Neuron Based on Nonlinear Polarization Rotation in a Semiconductor Optical Amplifier

Qiang Li,Zhi Wang,Huiying Wang,Can Cui,Chongqing Wu

DOI: https://doi.org/10.1016/j.optcom.2018.11.074

IF: 2.4

2018-01-01

Optics Communications

Abstract:We experimentally demonstrated an optoelectronic artificial neuron model, which can perform sigmoid transfer function. The proposed optoelectronic circuit can mimic the behavior of continuous sigmoidal neuron based on the nonlinear polarization rotation (NPR) in a single semiconductor optical amplifier (SOA). Because the current-induced NPR is anti-parallel to the optical power-induced NPR, excitatory and inhibitory stimuli to the proposed artificial neuron can be implemented by optical injection and electrical modulation, respectively. The response of the proposed neuron to excitatory inputs and inhibitory inputs are related to the initial state of polarization (SOP), bias current and power of probe beam. By adjusting the initial SOP, the neuron can be more likely to be excited or inhibited.
A Tunable Leaky Integrate-and-fire Neuron Based on One Neuromorphic Transistor and One Memristor

Huiwu Mao,Yixin Zhu,Shuo Ke,Ying Zhu,Kailu Shi,Xiangjing Wang,Changjin Wan,Qing Wan

DOI: https://doi.org/10.1063/5.0151312

IF: 4

2023-01-01

Applied Physics Letters

Abstract:Artificial leaky integrate-and-fire (LIF) neurons have attracted significant attention for building brain-like computing and neuromorphic systems. However, previous artificial LIF neurons have primarily focused on implementing integrate-and-fire function, and the function of dendritic modulation has rarely been reported. In this Letter, a tunable artificial LIF neuron based on an IGZO electric-double-layer (EDL) transistor and a TaOx memristor is fabricated, and dendritic modulation is investigated. An IGZO-based EDL transistor with a modulatory terminal is used to realize dendritic nonlinear integration and filtering capability, as well as the tunable neural excitability. An Ag/TaOx/ITO threshold switching memristor mimics the all-or-nothing spiking and threshold switching of the soma. By incorporating these two components in a customized way, such artificial LIF neuron can emulate the key function of biological neuron with rich computational flexibility. Our artificial LIF neurons with rich nonlinear dynamics have great potential to perform more complex tasks in future spiking neuromorphic systems.
Reconfigurable All-Optical Integrated Nonlinear Activator with Switchable Response Functions for Photonic Neural Networks

Bei Chen,Jian Wang,Zichao Zhao,Xiaowen Xiong,Jianyi Yang,Ming Li,Ninghua Zhu

DOI: https://doi.org/10.1364/ofc.2024.m4c.5

2024-01-01

Abstract:We experimentally demonstrate a reconfigurable all-optical integrated nonlinear activator with switchable response functions, including Gaussian, Radial Basis, Soft-plus, leaky ReLU, Swish and clamped ReLU functions, especially all triggered by low-power inputs.
The Optical-Electronic Integrated Spiking Neurons Based on Antiferroelectric Thin-Film Transistors

Xiaopeng Luo,Peng Yang,Shihao Yu,Xu Guo,Yefan Zhang,Yang Liu,Yi Sun,Yinan Wang,Sen Liu,Qingjiang Li

DOI: https://doi.org/10.1109/ted.2024.3450440

IF: 3.1

2024-09-28

IEEE Transactions on Electron Devices

Abstract:Sensory neurons are an important componentin achieving flexible and efficient intelligent sensing systems. Multimodal sensory neurons provide more comprehensive and reliable sensing than unimodal neurons, enabling them to better perform complex recognition or decision-making tasks. This article proposes a multimodal sensory neuron based on antiferroelectric thin-film transistors (AFeTFTs) for the first time. It can realize leaky integrate-and-fire (LIF) neuron behavior in both optical and electrical modes. The hafnium-based antiferroelectric capacitor is integrated on the gate of amorphous indium gallium zinc oxide thin-film transistors (a-IGZO, TFTs). By utilizing the polarization and depolarization properties of the antiferroelectric thin film Hf0.14Zr0.86O2, achieving the integration and firing behavior of LIF neurons in electrical mode. Additionally, based on the photosensitive properties of IGZO films, the device also realizes LIF neuron behavior in optical mode. The device exhibits reliable performance and high reproducibility. Since the IGZO film serves as both the channel layer of the transistor and the photosensitive unit of the neuron, the device has a compact structure and is convenient for integration. This work provides an effective approach for multimodal sensing fusion and functional integration, offering a solution for the realization of more efficient and flexible intelligent sensing systems.

engineering, electrical & electronic,physics, applied
Programmable low-threshold optical nonlinear activation functions for photonic neural networks

Ying Huang,Weiping Wang,Lei Qiao,Xiaoyan Hu,Tao Chu

DOI: https://doi.org/10.1364/OL.451287

IF: 3.6

2022-01-01

Optics Letters

Abstract:We experimentally demonstrate two types of programmable, low-threshold, optically controlled nonlinear activation functions, which are challenging to realize in photonic neural networks (PNNs). These devices rely on on-chip integrated Ge-Si photoelectric detectors and silicon electro-optical switches, and they generate rectified linear unit (ReLU) or sigmoid functions with arbitrary slopes without additional electrical processing. Both devices function at an extremely low threshold of 0.2mW. The embedding of these nonlinear activation functions into convolutional neural networks facilitates the attainment of high inference accuracies of up to 95% when applied to Modified National Institute of Standards and Technology (MNIST) handwritten digit-classification tasks. The devices are suitable for low-power PNNs with an arbitrary number of propagation layers in photonic-computing chips. (C) 2022 Optica Publishing Group
Optical Nonlinear Activation Functions Based on MZI-Structure for Optical Neural Networks

Qiang Li,Shengping Liu,Yang Zhao,Wei Wang,Ye Tian,Junbo Feng,Jin Guo

DOI: https://doi.org/10.1364/acpc.2020.t4d.4

2020-01-01

Abstract:We experimentally demonstrated an on-chip optical nonlinear activation function circuit for optical neural networks based on a conventional linear transformer, MZI-mesh. The proposed circuit is reconfigurable to perform multiple types of nonlinear activation functions.
Low-energy and tunable LIF neuron using SiGe bandgap-engineered resistive switching transistor

Yijoon Kim,Hyangwoo Kim,Kyounghwan Oh,Ju Hong Park,Byoung Don Kong,Chang-Ki Baek

DOI: https://doi.org/10.1186/s11671-024-04079-5

2024-08-23

Abstract:We have proposed leaky integrate-and-fire (LIF) neuron having low-energy consumption and tunable functionality without external circuit components. Our LIF neuron has a simple configuration consisting of only three components: one bandgap-engineered resistive switching transistor (BE-RST), one capacitor, and one resistor. Here, the crucial point is that BE-RST with a silicon-germanium heterojunction possesses an amplified hysteric current switching with a low latch-up voltage due to improved hole storage capability and impact ionization coefficient. Therefore, the proposed neuron utilizing BE-RST requires an energy consumption of 0.36 pJ/spike, which is approximately six times lower than 2.08 pJ/spike of pure silicon-RST based neuron. In addition, the spiking properties can be tuned by modulating the leakage rate and threshold through gate bias, which contributes to energy-efficient sparse-activity and high learning accuracy. As a result, our proposed neuron can be a promising candidate for executing various spiking neural network applications.
All‐Optical Nonlinear Neuron Based on Metallic Quantum Wells

Junru Niu,Hua Shao,Yiming Feng,Bingtao Gao,Yiyun Zhang,Ying Li,Hongsheng Chen,Haoliang Qian

DOI: https://doi.org/10.1002/adom.202300223

IF: 9

2023-04-25

Advanced Optical Materials

Abstract:An effectual all‐optical nonlinear neuron is proposed for implementing the sigmoid activation function in optical neural networks. The device uses a double‐ring resonator and metallic quantum wells exhibiting a large and ultrafast Kerr nonlinearity. It achieves a sigmoid activation function with a high modulation depth of 18.78 dB and a rapid transition rate of 13 GW cm−2 per 3 dB. Optical nonlinear neuron, as an essential implementation method for optical interconnection, could greatly promote the development of optical neuron networks, which appears to be a promising alternative to electronic neural networks that are limited by computing speed and the end of Moore's law. However, restricted by weak nonlinearities, the modulation performance of optical nonlinear neuron networks is much lower compared to their electronic counterparts. Therefore, an effectual all‐optical nonlinear neuron is proposed by combining a double‐ring resonator with field‐enhancement and metallic quantum wells (MQWs) with a large Kerr nonlinear response. Based on the large and ultrafast Kerr nonlinearity provided by MQWs, the device can achieve a modulation extinction ratio of 18.78 dB and a transition rate of 13 GW cm−2 per 3 dB. The work may provide a new route for integrated, ultrafast, and high‐efficient optical nonlinear neurons for on‐chip neuron networks.

materials science, multidisciplinary,optics
Optical ReLU-like Activation Function Based on a Semiconductor Laser with Optical Injection

Guanting Liu,Yiwei Shen,Ruiqian Li,Jingyi Yu,Xuming He,Cheng Wang

2023-11-02

Abstract:Artificial neural networks usually consist of successive linear multiply-accumulate operations and nonlinear activation functions. However, most optical neural networks only achieve the linear operation in the optical domain, while the optical implementation of activation function remains challenging. Here we present an optical ReLU-like activation function based on a semiconductor laser subject to the optical injection in experiment. The ReLU-like function is achieved in a broad regime above the Hopf bifurcation of the injection-locking diagram. In particular, the slope of the activation function is reconfigurable by tuning the frequency difference between the master laser and the slave laser.

Optics
Izhikevich-Inspired Optoelectronic Neurons with Excitatory and Inhibitory Inputs for Energy-Efficient Photonic Spiking Neural Networks

Yun-jhu Lee,Mehmet Berkay On,Xian Xiao,Roberto Proietti,S. J. Ben Yoo

DOI: https://doi.org/10.48550/arXiv.2105.02809

2021-05-03

Neural and Evolutionary Computing

Abstract:We designed, prototyped, and experimentally demonstrated, for the first time to our knowledge, an optoelectronic spiking neuron inspired by the Izhikevich model incorporating both excitatory and inhibitory optical spiking inputs and producing optical spiking outputs accordingly. The optoelectronic neurons consist of three transistors acting as electrical spiking circuits, a vertical-cavity surface-emitting laser (VCSEL) for optical spiking outputs, and two photodetectors for excitatory and inhibitory optical spiking inputs. Additional inclusion of capacitors and resistors complete the Izhikevich-inspired optoelectronic neurons, which receive excitatory and inhibitory optical spikes as inputs from other optoelectronic neurons. We developed a detailed optoelectronic neuron model in Verilog-A and simulated the circuit-level operation of various cases with excitatory input and inhibitory input signals. The experimental results closely resemble the simulated results and demonstrate how the excitatory inputs trigger the optical spiking outputs while the inhibitory inputs suppress the outputs. Utilizing the simulated neuron model, we conducted simulations using fully connected (FC) and convolutional neural networks (CNN). The simulation results using MNIST handwritten digits recognition show 90% accuracy on unsupervised learning and 97% accuracy on a supervised modified FC neural network. We further designed a nanoscale optoelectronic neuron utilizing quantum impedance conversion where a 200 aJ/spike input can trigger the output from on-chip nanolasers with 10 fJ/spike. The nanoscale neuron can support a fanout of ~80 or overcome 19 dB excess optical loss while running at 10 GSpikes/second in the neural network, which corresponds to 100x throughput and 1000x energy-efficiency improvement compared to state-of-art electrical neuromorphic hardware such as Loihi and NeuroGrid.
A Stacked Memristive Device Enabling Both Analog and Threshold Switching Behaviors for Artificial Leaky Integrate and Fire Neuron

Jingyao Bian,Ye Tao,Zhongqiang Wang,Xiaohan Zhang,Xiaoning Zhao,Ya Lin,Haiyang Xu,Yichun Liu

DOI: https://doi.org/10.1109/led.2022.3188786

IF: 4.8157

2022-09-03

IEEE Electron Device Letters

Abstract:Leaky integrate and fire (LIF) neurons are critical units for constructing a spiking neural network, in which neurons communicate with each other using spikes via synapses. Memristors, due to its specific nonlinear characteristics, are frequently introduced to emulate partial functions of LIF neurons for simplifying the circuit complexity, either the integration process or the fire action. Usually, a relatively complicated peripheral circuit needs to be engineered to assist the memristive device for complete emulation for biological neurons, which certainly would hinder the integration potential. Herein, we fabricated a stacked memristive device possessing both analog and threshold switching behaviors for constructing an artificial LIF neuron. Thus, the integration and fire functions were both accomplished within this single nanoscale device. In addition, the key neuronic functional of a biological neuron, including all-or-nothing spiking, threshold spiking, a refractory period, and strength-modulated frequency response were all successfully mimicked. The results demonstrate that the fabricated stacked memristor-based LIF neurons have great potential to construct high-density spiking neural network for neuromorphic computing.

engineering, electrical & electronic
Neuron-like Functionality of a Silicon Optical Transistor

Leo T. Varghese,Li Fan,Jian Wang,Yi Xuan,Andrew M. Weiner,Minghao Qi

DOI: https://doi.org/10.1364/cleo_at.2013.jtu4a.34

2013-01-01

Abstract:We demonstrate an all-silicon optical transistor that realizes NAND and NOR logic without the need to physically alter the device. The device is cascadable and may provide an efficient solid-state implementation of artificial neural networks.
Optical Implementation of Neural Learning Algorithms Based on Cross-gain Modulation in a Semiconductor Optical Amplifier

Qiang Li,Zhi Wang,Yansi Le,Chonghui Sun,Xiaojia Song,Chongqing Wu

DOI: https://doi.org/10.1117/12.2245976

2016-01-01

Abstract:Neuromorphic engineering has a wide range of applications in the fields of machine learning, pattern recognition, adaptive control, etc. Photonics, characterized by its high speed, wide bandwidth, low power consumption and massive parallelism, is an ideal way to realize ultrafast spiking neural networks (SNNs). Synaptic plasticity is believed to be critical for learning, memory and development in neural circuits. Experimental results have shown that changes of synapse are highly dependent on the relative timing of pre- and postsynaptic spikes. Synaptic plasticity in which presynaptic spikes preceding postsynaptic spikes results in strengthening, while the opposite timing results in weakening is called antisymmetric spike-timing-dependent plasticity (STDP) learning rule. And synaptic plasticity has the opposite effect under the same conditions is called antisymmetric anti-STDP learning rule. We proposed and experimentally demonstrated an optical implementation of neural learning algorithms, which can achieve both of antisymmetric STDP and anti-STDP learning rule, based on the cross-gain modulation (XGM) within a single semiconductor optical amplifier (SOA). The weight and height of the potentitation and depression window can be controlled by adjusting the injection current of the SOA, to mimic the biological antisymmetric STDP and anti-STDP learning rule more realistically. As the injection current increases, the width of depression and potentitation window decreases and height increases, due to the decreasing of recovery time and increasing of gain under a stronger injection current. Based on the demonstrated optical STDP circuit, ultrafast learning in optical SNNs can be realized.
Compact leak-integrate-fire neuron with auto-reset functionality based on a single spin–orbit torque magnetic tunnel junction device

Shiqi Wang,Runjie Chen,Chenyang Wang,Wenlong Cai,Daoqian Zhu,Ao Du,Zixi Wang,Zanhong Chen,Kewen Shi,Weisheng Zhao

DOI: https://doi.org/10.1063/5.0191119

IF: 4

2024-03-25

Applied Physics Letters

Abstract:Leaky-integrate-fire (LIF) neurons are core components to construct a spiking neural network. The emulation of LIF neurons has been implemented in spintronic devices, but typically suffers from challenges, such as relatively complex design and the requirement of additional operations for resetting. In this Letter, we propose a compact LIF neuron device realized within a single spin–orbit torque (SOT) magnetic tunnel junction device. Distinct from standard memory devices, the input SOT current for the integrating process is applied in a manner such that the magnetization cannot cross the hard plane. Consequently, the device can automatically reset to its original state by the combined effects of anisotropy and damping, which play a vital role during the leaky process as well. We verify the proposal in three types of SOT devices by micromagnetic simulations, and the power consumption is estimated as 0.1 pJ/spike. The auto-reset process is further captured by our single-shot dynamic experiments. With the state-of-the-art SOT technology, our work provides a concise and plausible scheme to mimic LIF neurons, which is of practical interest for neuromorphic computing.

physics, applied
Capacitor-less Stochastic Leaky-FeFET Neuron of Both Excitatory and Inhibitory Connections for SNN with Reduced Hardware Cost

Jin Luo,Liutao Yu,Tianyi Liu,Mengxuan Yang,Zhiyuan Fu,Zhongxin Liang,Liang Chen,Cheng Chen,Shuhan Liu,Si Wu,Qianqian Huang,Ru Huang

DOI: https://doi.org/10.1109/iedm19573.2019.8993535

2019-01-01

Abstract:For the unsupervised learning of spiking neural networks (SNNs), both excitatory and inhibitory signals with different arriving time should be connected to the artificial neurons, causing the high-hardware-cost issue for the implementation. In this work, a capacitor-less neuron based on the novel Leaky-FeFET (L-FeFET) device is proposed and experimentally demonstrated with remarkably reduced hardware cost of only two transistors and one resistor. For the first time, by exploiting the physics of directional ferroelectric polarization switching, both excitatory and inhibitory input connections can be emulated for the biomimetic neuronal dynamics in both experiments and simulations. Moreover, the neuronal stochastic behavior can be realized without additional hardware cost due to the inherent ferroelectric dynamics. Based on the proposed neuron, a bio-plausible SNN is efficiently implemented for clustering and inference tasks. Both biomimetic clustering by self-organizing map learning and high-accuracy (92%) inference by winner-take-all learning are realized. This work provides a promising highly-integrated neuron solution for brain-inspired neuromorphic computing system.
Bio-Inspired Neurons Based on Novel Leaky-FeFET with Ultra-Low Hardware Cost and Advanced Functionality for All-Ferroelectric Neural Network

C. Chen,M. Yang,S. Liu,T. Liu,K. Zhu,Y. Zhao,H. Wang,Q. Huang,R. Huang

DOI: https://doi.org/10.23919/vlsit.2019.8776495

2019-01-01

Abstract:For the brain-inspired neuromorphic computing, various emerging memory devices, including FeFET, have been applied to develop the artificial synapses, while the artificial neurons are still mostly CMOS-implemented and suffer from high-hardware-cost issue, especially when expanding advanced functions. In this work, a novel leaky-FeFET (L-FeFET) based on partially crystallized Hf 0.5 Zr 0.5 O 2 layer is designed to mimic biological neurons. For the first time, we propose and experimentally demonstrate a capacitor-less L-FeFET neuron for basic leaky-integrate-and-fire function with ultra-low hardware cost of only one transistor and one resistor. Furthermore, a new hybrid L-FeFET-CMOS neuron is implemented to expand advanced spike-frequency adaption with almost half of hardware cost compared with CMOS neuron. This work provides a highly-integrated and inherently-low-energy implementation for neuron and the possibility for all-ferroelectric neural networks.
Optimization of Leaky Integrate-and-Fire Neuron Circuits Based on Nanoporous Graphene Memristors

Kannan Udaya Mohanan,Seyed Mehdi Sattari-Esfahlan,Eou-Sik Cho,Chang-Hyun Kim

DOI: https://doi.org/10.1109/jeds.2024.3352827

2024-02-03

IEEE Journal of the Electron Devices Society

Abstract:Artificial neurons form the core of neuromorphic computing which is emerging as an alternative for the von Neumann computing architecture. However, existing neuron architectures still lack in area efficiency, especially considering the huge size of modern neural networks requiring millions of neurons. Here, we report on a compact leaky integrate and fire (LIF) neuron circuit based on graphene memristor device. The LIF circuit exhibits various biological properties like threshold control, leaky integration and reset behavior. Circuit parameters like the synaptic resistance and membrane capacitance act as additional control parameters whereby the spike frequency of the circuit can be effectively controlled. Uniquely, the circuit exhibits biologically realistic frequencies as low as 286 Hz. The results suggest the suitability of this compact and biorealistic LIF neuron circuit towards future bioinspired computing systems

engineering, electrical & electronic
Leaky integrate-and-fire neurons based on perovskite memristor for spiking neural networks

Jia-Qin Yang,Ruopeng Wang,Zhan-Peng Wang,Qin-Yuan Ma,Jing-Yu Mao,Yi Ren,Xiaoyang Yang,Ye Zhou,Su-Ting Han

DOI: https://doi.org/10.1016/j.nanoen.2020.104828

IF: 17.6

2020-08-01

Nano Energy

Abstract:Artificial neuron is an important part of constructing neuromorphic network in which information can be computed with high parallelism and efficiency like in the human brain. However, owing to the poor biological plausibility, artificial neurons based on traditional complementary metal-oxide-semiconductor (CMOS) platform fail to reveal the rich ion dynamics of the biological counterparts. Organic–inorganic halide perovskites (OHPs) are prospective for imitating the ion dynamics on the membranes of biological neurons because of its intrinsic ion migration. Herein, a diffusive CH3NH3PbI3(MAPbI3)-based memristor with superior amplitude-frequency characteristics and highly linear conductivity modulation for more than 1000 states have been fabricated for the construction of a leaky integrate-and-fire (LIF) bio-inspired neuron. The as-designed LIF model can successfully emulate the leakage, spatiotemporal integration and firing functions in a biological neuron. Moreover, by connecting LIF neurons with a 2*2 non-volatile Al2O3-based synaptic array, a simple spiking neural network (SNN) which is called the 3rd generation of neural network has been implemented at the hardware level to study the cognitive performance of the network. The SNN exhibits outstanding selective sensitivity to particular input sequence, indicating the excellent adaptability and versatility of the network for future applications of neuromorphic computing by utilizing novel ionotropic device.

materials science, multidisciplinary,chemistry, physical,physics, applied,nanoscience & nanotechnology
A Monolayer Leaky Integrate‐and‐Fire Neuron for 2D Memristive Neuromorphic Networks

Song Hao,Xinglong Ji,Shuai Zhong,Khin Yin Pang,Kian Guan Lim,Chong,Rong Zhao

DOI: https://doi.org/10.1002/aelm.201901335

IF: 6.2

2020-01-01

Advanced Electronic Materials

Abstract:2D material based memristors have exhibited superior performance as artificial synapses for neuromorphic computing. However, 2D artificial neurons as have note been exploited as an indispensable computational element owing to the rich dynamics of neurons, which impede the construction of a 2D neuromorphic network. A methodology is developed by introducing ionic migration dynamics and electrochemical reaction into monolayer MoS2 single crystal and a 2D artificial neuron is realized. The sophisticated electrophysiology process of leaky integrate‐and‐fire (LIF) is emulated by the injection and extraction of Ag+ ions under an e‐field in a monolayer MoS2 device with fine‐tuned channel length. Moreover, the fire frequency and relaxation time of the artificial neurons can be readily modulated by adjusting the input voltage pulses. By directly capturing conductive filament under a scanning electron microscope, the underlying mechanism of the unique resistive switching of the 2D artificial neuron is attributed to the rapid diffusion and migration of Ag in MoS2 lattice under the e‐field. The formation and rupture of the conductive Ag filament enable volatile resistive switching and LIF behaviors. Furthermore, MoS2‐based neurons are integrated with a nonvolatile synapse array to build a full memristive artificial neural network and implement pattern classification, paving the way for the construction of 2D neuromorphic networks.
GLIF: A Unified Gated Leaky Integrate-and-Fire Neuron for Spiking Neural Networks

Xingting Yao,Fanrong Li,Zitao Mo,Jian Cheng

DOI: https://doi.org/10.48550/arXiv.2210.13768

2023-02-14

Abstract:Spiking Neural Networks (SNNs) have been studied over decades to incorporate their biological plausibility and leverage their promising energy efficiency. Throughout existing SNNs, the leaky integrate-and-fire (LIF) model is commonly adopted to formulate the spiking neuron and evolves into numerous variants with different biological features. However, most LIF-based neurons support only single biological feature in different neuronal behaviors, limiting their expressiveness and neuronal dynamic diversity. In this paper, we propose GLIF, a unified spiking neuron, to fuse different bio-features in different neuronal behaviors, enlarging the representation space of spiking neurons. In GLIF, gating factors, which are exploited to determine the proportion of the fused bio-features, are learnable during training. Combining all learnable membrane-related parameters, our method can make spiking neurons different and constantly changing, thus increasing the heterogeneity and adaptivity of spiking neurons. Extensive experiments on a variety of datasets demonstrate that our method obtains superior performance compared with other SNNs by simply changing their neuronal formulations to GLIF. In particular, we train a spiking ResNet-19 with GLIF and achieve $77.35\%$ top-1 accuracy with six time steps on CIFAR-100, which has advanced the state-of-the-art. Codes are available at \url{<a class="link-external link-https" href="https://github.com/Ikarosy/Gated-LIF" rel="external noopener nofollow">this https URL</a>}.

Neural and Evolutionary Computing,Artificial Intelligence,Computer Vision and Pattern Recognition

All-optical Leaky-integrate-and-fire Neuron with Half-sigmoid Transfer Function

Artificial Neuron Based on Nonlinear Polarization Rotation in a Semiconductor Optical Amplifier

A Tunable Leaky Integrate-and-fire Neuron Based on One Neuromorphic Transistor and One Memristor

Reconfigurable All-Optical Integrated Nonlinear Activator with Switchable Response Functions for Photonic Neural Networks

The Optical-Electronic Integrated Spiking Neurons Based on Antiferroelectric Thin-Film Transistors

Programmable low-threshold optical nonlinear activation functions for photonic neural networks

Optical Nonlinear Activation Functions Based on MZI-Structure for Optical Neural Networks

Low-energy and tunable LIF neuron using SiGe bandgap-engineered resistive switching transistor

All‐Optical Nonlinear Neuron Based on Metallic Quantum Wells

Optical ReLU-like Activation Function Based on a Semiconductor Laser with Optical Injection

Izhikevich-Inspired Optoelectronic Neurons with Excitatory and Inhibitory Inputs for Energy-Efficient Photonic Spiking Neural Networks

A Stacked Memristive Device Enabling Both Analog and Threshold Switching Behaviors for Artificial Leaky Integrate and Fire Neuron

Neuron-like Functionality of a Silicon Optical Transistor

Optical Implementation of Neural Learning Algorithms Based on Cross-gain Modulation in a Semiconductor Optical Amplifier

Compact leak-integrate-fire neuron with auto-reset functionality based on a single spin–orbit torque magnetic tunnel junction device

Capacitor-less Stochastic Leaky-FeFET Neuron of Both Excitatory and Inhibitory Connections for SNN with Reduced Hardware Cost

Bio-Inspired Neurons Based on Novel Leaky-FeFET with Ultra-Low Hardware Cost and Advanced Functionality for All-Ferroelectric Neural Network

Optimization of Leaky Integrate-and-Fire Neuron Circuits Based on Nanoporous Graphene Memristors

Leaky integrate-and-fire neurons based on perovskite memristor for spiking neural networks

A Monolayer Leaky Integrate‐and‐Fire Neuron for 2D Memristive Neuromorphic Networks

GLIF: A Unified Gated Leaky Integrate-and-Fire Neuron for Spiking Neural Networks