Spiking neural networks(SNNs)represent a biologically-inspired computational framework that bridges neuroscience and artificial intelligence,offering unique advantages in temporal data processing,energy efficiency,and...Spiking neural networks(SNNs)represent a biologically-inspired computational framework that bridges neuroscience and artificial intelligence,offering unique advantages in temporal data processing,energy efficiency,and real-time decision-making.This paper explores the evolution of SNN technologies,emphasizing their integration with advanced learning mechanisms such as spike-timing-dependent plasticity(STDP)and hybridization with deep learning architectures.Leveraging memristors as nanoscale synaptic devices,we demonstrate significant enhancements in energy efficiency,adaptability,and scalability,addressing key challenges in neuromorphic computing.Through phase portraits and nonlinear dynamics analysis,we validate the system’s stability and robustness under diverse workloads.These advancements position SNNs as a transformative technology for applications in robotics,IoT,and adaptive low-power AI systems,paving the way for future innovations in neuromorphic hardware and hybrid learning paradigms.展开更多
Spiking neural networks(SNN)represent a paradigm shift toward discrete,event-driven neural computation that mirrors biological brain mechanisms.This survey systematically examines current SNN research,focusing on trai...Spiking neural networks(SNN)represent a paradigm shift toward discrete,event-driven neural computation that mirrors biological brain mechanisms.This survey systematically examines current SNN research,focusing on training methodologies,hardware implementations,and practical applications.We analyze four major training paradigms:ANN-to-SNN conversion,direct gradient-based training,spike-timing-dependent plasticity(STDP),and hybrid approaches.Our review encompasses major specialized hardware platforms:Intel Loihi,IBM TrueNorth,SpiNNaker,and BrainScaleS,analyzing their capabilities and constraints.We survey applications spanning computer vision,robotics,edge computing,and brain-computer interfaces,identifying where SNN provide compelling advantages.Our comparative analysis reveals SNN offer significant energy efficiency improvements(1000-10000×reduction)and natural temporal processing,while facing challenges in scalability and training complexity.We identify critical research directions including improved gradient estimation,standardized benchmarking protocols,and hardware-software co-design approaches.This survey provides researchers and practitioners with a comprehensive understanding of current SNN capabilities,limitations,and future prospects.展开更多
Spiking Neural Network(SNN)inspired by the biological triggering mechanism of neurons to provide a novel solution for plant disease detection,offering enhanced performance and efficiency in contrast to Artificial Neur...Spiking Neural Network(SNN)inspired by the biological triggering mechanism of neurons to provide a novel solution for plant disease detection,offering enhanced performance and efficiency in contrast to Artificial Neural Networks(ANN).Unlike conventional ANNs,which process static images without fully capturing the inherent temporal dynamics,our approach represents the first implementation of SNNs tailored explicitly for agricultural disease classification,integrating an encoding method to convert static RGB plant images into temporally encoded spike trains.Additionally,while Bernoulli trials and standard deep learning architectures likeConvolutionalNeuralNetworks(CNNs)and Fully Connected Neural Networks(FCNNs)have been used extensively,our work is the first to integrate these trials within an SNN framework specifically for agricultural applications.This integration not only refines spike regulation and reduces computational overhead by 30%but also delivers superior accuracy(93.4%)in plant disease classification,marking a significant advancement in precision agriculture that has not been previously explored.Our approach uniquely transforms static plant leaf images into time-dependent representations,leveraging SNNs’intrinsic temporal processing capabilities.This approach aligns with the inherent ability of SNNs to capture dynamic,timedependent patterns,making them more suitable for detecting disease activations in plants than conventional ANNs that treat inputs as static entities.Unlike prior works,our hybrid encoding scheme dynamically adapts to pixel intensity variations(via threshold),enabling robust feature extraction under diverse agricultural conditions.The dual-stage preprocessing customizes the SNN’s behavior in two ways:the encoding threshold is derived from pixel distributions in diseased regions,and Bernoulli trials selectively reduce redundant spikes to ensure energy efficiency on low-power devices.We used a comprehensive dataset of 87,000 RGB images of plant leaves,which included 38 distinct classes of healthy and unhealthy leaves.To train and evaluate three distinct neural network architectures,DeepSNN,SimpleCNN,and SimpleFCNN,the dataset was rigorously preprocessed,including stochastic rotation,horizontal flip,resizing,and normalization.Moreover,by integrating Bernoulli trials to regulate spike generation,ourmethod focuses on extracting themost relevant featureswhile reducingcomputational overhead.Using a comprehensivedatasetof87,000RGB images across 38 classes,we rigorously preprocessed the data and evaluated three architectures:DeepSNN,SimpleCNN,and SimpleFCNN.The results demonstrate that DeepSNN outperforms the other models,achieving superior accuracy,efficient feature extraction,and robust spike management,thereby establishing the potential of SNNs for real-time,energy-efficient agricultural applications.展开更多
Neuromorphic photonic computing has emerged as a competitive computing paradigm to overcome the bottlenecks of the von-Neumann architecture.Linear weighting and nonlinear spike activation are two fundamental functions...Neuromorphic photonic computing has emerged as a competitive computing paradigm to overcome the bottlenecks of the von-Neumann architecture.Linear weighting and nonlinear spike activation are two fundamental functions of a photonic spiking neural network(PSNN).However,they are separately implemented with different photonic materials and devices,hindering the large-scale integration of PSNN.Here,we propose,fabricate and experimentally demonstrate a photonic neuro-synaptic chip enabling the simultaneous implementation of linear weighting and nonlinear spike activation based on a distributed feedback(DFB)laser with a saturable absorber(DFB-SA).A prototypical system is experimentally constructed to demonstrate the parallel weighted function and nonlinear spike activation.Furthermore,a fourchannel DFB-SA laser array is fabricated for realizing matrix convolution of a spiking convolutional neural network,achieving a recognition accuracy of 87%for the MNIST dataset.The fabricated neuro-synaptic chip offers a fundamental building block to construct the large-scale integrated PSNN chip.展开更多
For accelerating the supervised learning by the SpikeProp algorithm with the temporal coding paradigm in spiking neural networks (SNNs), three learning rate adaptation methods (heuristic rule, delta-delta rule, and de...For accelerating the supervised learning by the SpikeProp algorithm with the temporal coding paradigm in spiking neural networks (SNNs), three learning rate adaptation methods (heuristic rule, delta-delta rule, and delta-bar-delta rule), which are used to speed up training in artificial neural networks, are used to develop the training algorithms for feedforward SNN. The performance of these algorithms is investigated by four experiments: classical XOR (exclusive or) problem, Iris dataset, fault diagnosis in the Tennessee Eastman process, and Poisson trains of discrete spikes. The results demonstrate that all the three learning rate adaptation methods are able to speed up convergence of SNN compared with the original SpikeProp algorithm. Furthermore, if the adaptive learning rate is used in combination with the momentum term, the two modifications will balance each other in a beneficial way to accomplish rapid and steady convergence. In the three learning rate adaptation methods, delta-bar-delta rule performs the best. The delta-bar-delta method with momentum has the fastest convergence rate, the greatest stability of training process, and the maximum accuracy of network learning. The proposed algorithms in this paper are simple and efficient, and consequently valuable for practical applications of SNN.展开更多
Artificial neural networks(ANNs)have led to landmark changes in many fields,but they still differ significantly fromthemechanisms of real biological neural networks and face problems such as high computing costs,exces...Artificial neural networks(ANNs)have led to landmark changes in many fields,but they still differ significantly fromthemechanisms of real biological neural networks and face problems such as high computing costs,excessive computing power,and so on.Spiking neural networks(SNNs)provide a new approach combined with brain-like science to improve the computational energy efficiency,computational architecture,and biological credibility of current deep learning applications.In the early stage of development,its poor performance hindered the application of SNNs in real-world scenarios.In recent years,SNNs have made great progress in computational performance and practicability compared with the earlier research results,and are continuously producing significant results.Although there are already many pieces of literature on SNNs,there is still a lack of comprehensive review on SNNs from the perspective of improving performance and practicality as well as incorporating the latest research results.Starting from this issue,this paper elaborates on SNNs along the complete usage process of SNNs including network construction,data processing,model training,development,and deployment,aiming to provide more comprehensive and practical guidance to promote the development of SNNs.Therefore,the connotation and development status of SNNcomputing is reviewed systematically and comprehensively from four aspects:composition structure,data set,learning algorithm,software/hardware development platform.Then the development characteristics of SNNs in intelligent computing are summarized,the current challenges of SNNs are discussed and the future development directions are also prospected.Our research shows that in the fields of machine learning and intelligent computing,SNNs have comparable network scale and performance to ANNs and the ability to challenge large datasets and a variety of tasks.The advantages of SNNs over ANNs in terms of energy efficiency and spatial-temporal data processing have been more fully exploited.And the development of programming and deployment tools has lowered the threshold for the use of SNNs.SNNs show a broad development prospect for brain-like computing.展开更多
Spiking neural networks(SNNs)utilize brain-like spatiotemporal spike encoding for simulating brain functions.Photonic SNN offers an ultrahigh speed and power efficiency platform for implementing high-performance neuro...Spiking neural networks(SNNs)utilize brain-like spatiotemporal spike encoding for simulating brain functions.Photonic SNN offers an ultrahigh speed and power efficiency platform for implementing high-performance neuromorphic computing.Here,we proposed a multi-synaptic photonic SNN,combining the modified remote supervised learning with delayweight co-training to achieve pattern classification.The impact of multi-synaptic connections and the robustness of the network were investigated through numerical simulations.In addition,the collaborative computing of algorithm and hardware was demonstrated based on a fabricated integrated distributed feedback laser with a saturable absorber(DFB-SA),where 10 different noisy digital patterns were successfully classified.A functional photonic SNN that far exceeds the scale limit of hardware integration was achieved based on time-division multiplexing,demonstrating the capability of hardware-algorithm co-computation.展开更多
Spiking Neural Network is known as the third-generation artificial neural network whose development has great potential.With the help of Spike Layer Error Reassignment in Time for error back-propagation,this work pres...Spiking Neural Network is known as the third-generation artificial neural network whose development has great potential.With the help of Spike Layer Error Reassignment in Time for error back-propagation,this work presents a new network called SpikeGoogle,which is implemented with GoogLeNet-like inception module.In this inception module,different convolution kernels and max-pooling layer are included to capture deep features across diverse scales.Experiment results on small NMNIST dataset verify the results of the authors’proposed SpikeGoogle,which outperforms the previous Spiking Convolutional Neural Network method by a large margin.展开更多
In recent years, spiking neural networks(SNNs) have received increasing attention of research in the field of artificial intelligence due to their high biological plausibility, low energy consumption, and abundant spa...In recent years, spiking neural networks(SNNs) have received increasing attention of research in the field of artificial intelligence due to their high biological plausibility, low energy consumption, and abundant spatio-temporal information.However, the non-differential spike activity makes SNNs more difficult to train in supervised training. Most existing methods focusing on introducing an approximated derivative to replace it, while they are often based on static surrogate functions. In this paper, we propose a progressive surrogate gradient learning for backpropagation of SNNs, which is able to approximate the step function gradually and to reduce information loss. Furthermore, memristor cross arrays are used for speeding up calculation and reducing system energy consumption for their hardware advantage. The proposed algorithm is evaluated on both static and neuromorphic datasets using fully connected and convolutional network architecture, and the experimental results indicate that our approach has a high performance compared with previous research.展开更多
Spiking neural networks(SNNs) are widely used in many fields because they work closer to biological neurons.However,due to its computational complexity,many SNNs implementations are limited to computer programs.First,...Spiking neural networks(SNNs) are widely used in many fields because they work closer to biological neurons.However,due to its computational complexity,many SNNs implementations are limited to computer programs.First,this paper proposes a multi-synaptic circuit(MSC) based on memristor,which realizes the multi-synapse connection between neurons and the multi-delay transmission of pulse signals.The synapse circuit participates in the calculation of the network while transmitting the pulse signal,and completes the complex calculations on the software with hardware.Secondly,a new spiking neuron circuit based on the leaky integrate-and-fire(LIF) model is designed in this paper.The amplitude and width of the pulse emitted by the spiking neuron circuit can be adjusted as required.The combination of spiking neuron circuit and MSC forms the multi-synaptic spiking neuron(MSSN).The MSSN was simulated in PSPICE and the expected result was obtained,which verified the feasibility of the circuit.Finally,a small SNN was designed based on the mathematical model of MSSN.After the SNN is trained and optimized,it obtains a good accuracy in the classification of the IRIS-dataset,which verifies the practicability of the design in the network.展开更多
Adolescent Idiopathic Scoliosis(AIS)is a deformity of the spine that affects teenagers.The current method for detecting AIS is based on radiographic images which may increase the risk of cancer growth due to radiation...Adolescent Idiopathic Scoliosis(AIS)is a deformity of the spine that affects teenagers.The current method for detecting AIS is based on radiographic images which may increase the risk of cancer growth due to radiation.Photogrammetry is another alternative used to identify AIS by distinguishing the curves of the spine from the surface of a human’s back.Currently,detecting the curve of the spine is manually performed,making it a time-consuming task.To overcome this issue,it is crucial to develop a better model that automatically detects the curve of the spine and classify the types of AIS.This research proposes a new integration of ESNN and Feature Extraction(FE)methods and explores the architecture of ESNN for the AIS classification model.This research identifies the optimal Feature Extraction(FE)methods to reduce computational complexity.The ability of ESNN to provide a fast result with a simplicity and performance capability makes this model suitable to be implemented in a clinical setting where a quick result is crucial.A comparison between the conventional classifier(Support Vector Machine(SVM),Multi-layer Perceptron(MLP)and Random Forest(RF))with the proposed AIS model also be performed on a dataset collected by an orthopedic expert from Hospital Universiti Kebangsaan Malaysia(HUKM).This dataset consists of various photogrammetry images of the human back with different types ofMalaysian AIS patients to solve the scoliosis problem.The process begins by pre-processing the images which includes resizing and converting the captured pictures to gray-scale images.This is then followed by feature extraction,normalization,and classification.The experimental results indicate that the integration of LBP and ESNN achieves higher accuracy compared to the performance of multiple baseline state-of-the-art Machine Learning for AIS classification.This demonstrates the capability of ESNN in classifying the types of AIS based on photogrammetry images.展开更多
Medical image classification becomes a vital part of the design of computer aided diagnosis(CAD)models.The conventional CAD models are majorly dependent upon the shapes,colors,and/or textures that are problem oriented...Medical image classification becomes a vital part of the design of computer aided diagnosis(CAD)models.The conventional CAD models are majorly dependent upon the shapes,colors,and/or textures that are problem oriented and exhibited complementary in medical images.The recently developed deep learning(DL)approaches pave an efficient method of constructing dedicated models for classification problems.But the maximum resolution of medical images and small datasets,DL models are facing the issues of increased computation cost.In this aspect,this paper presents a deep convolutional neural network with hierarchical spiking neural network(DCNN-HSNN)for medical image classification.The proposed DCNN-HSNN technique aims to detect and classify the existence of diseases using medical images.In addition,region growing segmentation technique is involved to determine the infected regions in the medical image.Moreover,NADAM optimizer with DCNN based Capsule Network(CapsNet)approach is used for feature extraction and derived a collection of feature vectors.Furthermore,the shark smell optimization algorithm(SSA)based HSNN approach is utilized for classification process.In order to validate the better performance of the DCNN-HSNN technique,a wide range of simulations take place against HIS2828 and ISIC2017 datasets.The experimental results highlighted the effectiveness of the DCNN-HSNN technique over the recent techniques interms of different measures.Please type your abstract here.展开更多
Deep reinforcement learning(DRL)achieves success through the representational capabilities of deep neural networks(DNNs).Compared to DNNs,spiking neural networks(SNNs),known for their binary spike information processi...Deep reinforcement learning(DRL)achieves success through the representational capabilities of deep neural networks(DNNs).Compared to DNNs,spiking neural networks(SNNs),known for their binary spike information processing,exhibit more biological characteristics.However,the challenge of using SNNs to simulate more biologically characteristic neuronal dynamics to optimize decision-making tasks remains,directly related to the information integration and transmission in SNNs.Inspired by the advanced computational power of dendrites in biological neurons,we propose a multi-dendrite spiking neuron(MDSN)model based on Multi-compartment spiking neurons(MCN),expanding dendrite types from two to multiple and deriving the analytical solution of somatic membrane potential.We apply the MDSN to deep distributional reinforcement learning to enhance its performance in executing complex decisionmaking tasks.The proposed model can effectively and adaptively integrate and transmit meaningful information from different sources.Our model uses a bioinspired event-enhanced dendrite structure to emphasize features.Meanwhile,by utilizing dynamic membrane potential thresholds,it adaptively maintains the homeostasis of MDSN.Extensive experiments on Atari games show that the proposed model outperforms some state-of-the-art spiking distributional RL models by a significant margin.展开更多
While the exponential growth of the space sector and new operative concepts ask for higher spacecraft autonomy,the development of AI-assisted space systems was so far hindered by the low availability of power and ener...While the exponential growth of the space sector and new operative concepts ask for higher spacecraft autonomy,the development of AI-assisted space systems was so far hindered by the low availability of power and energy typical of space applications.In this context,Spiking Neural Networks(SNN)are highly attractive because of their theoretically superior energy efficiency due to their inherently sparse activity induced by neurons communicating by means of binary spikes.Nevertheless,the ability of SNN to reach such efficiency on real world tasks is still to be demonstrated in practice.To evaluate the feasibility of utilizing SNN onboard spacecraft,this work presents a numerical analysis and comparison of different SNN techniques applied to scene classification for the EuroSAT dataset.Such tasks are of primary importance for space applications and constitute a valuable test case given the abundance of competitive methods available to establish a benchmark.Particular emphasis is placed on models based on temporal coding,where crucial information is encoded in the timing of neuron spikes.These models promise even greater efficiency of resulting networks,as they maximize the sparsity properties inherent in SNN.A reliable metric capable of comparing different architectures in a hardware-agnostic way is developed to establish a clear theoretical dependence between architecture parameters and the energy consumption that can be expected onboard the spacecraft.The potential of this novel method and its flexibility to describe specific hardware platforms is demonstrated by its application to predicting the energy consumption of a BrainChip Akida AKD1000 neuromorphic processor.展开更多
Photonic hardware implementation of spiking neural networks,regarded as a viable potential paradigm for ultra-high speed and energy efficiency computing,leverages spatiotemporal spike encoding and event-driven dynamic...Photonic hardware implementation of spiking neural networks,regarded as a viable potential paradigm for ultra-high speed and energy efficiency computing,leverages spatiotemporal spike encoding and event-driven dynamics to simulate brain-like parallel information processing.Silicon-based microring resonators(MRRs)offer a power efficiency and ultrahigh flexibility scheme to mimic biological neuron,however,their substantial potential for integrated neuromorphic systems remains limited by insufficient exploration of MRR-based spiking digital and analog computation.Here,an all-optical neural dynamics framework,encompassing both excitatory and inhibitory behaviors based on multi-wavelength auxiliary and competition mechanism in an MRR,is proposed numerically.Leveraging multi-wavelength resonance characteristics and wavelength division multiplexing(WDM)technology,a single MRR implements the five fundamental optical digital logic gates:AND,OR,NOT,XNOR and XOR.Besides,the cascading capabilities of MRR-based spiking neurons are demonstrated through multi-level digital logic gates including NAND,NOR,4-input AND,8-input AND,and a full adder,emphasizing their promise for large-scale digital logic networks.Furthermore,an exemplary binary convolution has been achieved by utilizing the proposed MRR-based digital logic operation,illustrating the potential of all-optical binary convolution to compute image gradient magnitudes for edge detection.Such passive photonic neurons and networks promise access to the high transmission speed and low power consumption inherent to optical systems,thus enabling direct hardware-algorithm co-computation and accelerating artificial intelligence.展开更多
Diffusive threshold switching(TS)memristors have emerged as a promising candidate for artificial neurons,effectively replicating neuronal functions and enabling spiking neural networks(SNNs)to emulate the low-power pr...Diffusive threshold switching(TS)memristors have emerged as a promising candidate for artificial neurons,effectively replicating neuronal functions and enabling spiking neural networks(SNNs)to emulate the low-power processing of biological brains.In this study,we present an artificial neuron based on a Pt/Ag/ZnO/Pt volatile memristor,which exhibits exceptional TS characteristics,including electro-forming-free operation,low voltage requirements(<0.2 V),high stability(2.25%variation over 1024 cycles),a high on/off ratio(106),and inherent self-compliance.These Pt/Ag/ZnO/Pt diffusive memristors are employed to simultaneously emulate oscillation neurons and leaky integrate-and-fire(LIF)neurons,enabling precise modulation of oscillation and firing frequencies through pulse parameters while maintaining low energy consumption(1.442 nJ per spike).We further integrate the oscillation and LIF neurons as input and output neurons,respectively,in a two-layer SNN,achieving a high classification accuracy of 89.17%on MNIST-based voltage images.This work underscores the potential of ZnO diffusive memristors in emulating stable artificial neurons and highlights their promise for advanced neuromorphic computing applications using SNNs.展开更多
Spike-based neural networks,which use spikes or action potentialsto represent information,have gained a lot of attention because of their high energyefficiency and low power consumption.To fully leverage its advantage...Spike-based neural networks,which use spikes or action potentialsto represent information,have gained a lot of attention because of their high energyefficiency and low power consumption.To fully leverage its advantages,convertingthe external analog signals to spikes is an essential prerequisite.Conventionalapproaches including analog-to-digital converters or ring oscillators,and sensorssuffer from high power and area costs.Recent efforts are devoted to constructingartificial sensory neurons based on emerging devices inspired by the biologicalsensory system.They can simultaneously perform sensing and spike conversion,overcoming the deficiencies of traditional sensory systems.This review summarizesand benchmarks the recent progress of artificial sensory neurons.It starts with thepresentation of various mechanisms of biological signal transduction,followed bythe systematic introduction of the emerging devices employed for artificial sensoryneurons.Furthermore,the implementations with different perceptual capabilitiesare briefly outlined and the key metrics and potential applications are also provided.Finally,we highlight the challenges and perspectives for the future development of artificial sensory neurons.展开更多
Significant progress has been made in brain-computer science and technology through applying spiking neural networks(SNNs)[1].More recently,due to its potential of processing complex spatio-temporal information,SNNs h...Significant progress has been made in brain-computer science and technology through applying spiking neural networks(SNNs)[1].More recently,due to its potential of processing complex spatio-temporal information,SNNs have been successfully applied in many fields,such as action recognition[2].There are two effective ways to design network models:converting artificial neural networks(ANNs)into SNNs and directly designing SNNs based on spike mechanisms.In the ANN-SNN method,the integrate-andfire(IF)neurons are used to replace the activation layer to convert ANNs into SNNs,which have some inherent drawbacks,such as inevitable accuracy loss,more delays and energy consumption.Although existing direct training strategies have shown outstanding performance in image classification tasks,SNNs face significant difficulties in handling complex video understanding tasks.In light of the considerable success achieved by the ANNs employed in the field of human action recognition,more researchers have recently focused their attention on using SNNs for action recognition.In order to design more efficient SNNs,some researchers have proposed a series of effective training and feature learning mechanisms in residual network,e.g.,Hu et al.展开更多
The rapid advancement of deep learning and the emergence of largescale neural models,such as bidirectional encoder representations from transformers(BERT),generative pre-trained transformer(GPT),and large language mod...The rapid advancement of deep learning and the emergence of largescale neural models,such as bidirectional encoder representations from transformers(BERT),generative pre-trained transformer(GPT),and large language model Meta AI(LLaMa),have brought significant computational and energy challenges.Neuromorphic computing presents a biologically inspired approach to addressing these issues,leveraging event-driven processing and in-memory computation for enhanced energy efficiency.This survey explores the intersection of neuromorphic computing and large-scale deep learning models,focusing on neuromorphic models,learning methods,and hardware.We highlight transferable techniques from deep learning to neuromorphic computing and examine the memoryrelated scalability limitations of current neuromorphic systems.Furthermore,we identify potential directions to enable neuromorphic systems to meet the growing demands of modern AI workloads.展开更多
Artificial neural networks(ANNs)have made great strides in the field of remote sensing image object detection.However,low detection efficiency and high power consumption have always been significant bottlenecks in rem...Artificial neural networks(ANNs)have made great strides in the field of remote sensing image object detection.However,low detection efficiency and high power consumption have always been significant bottlenecks in remote sensing.Spiking neural networks(SNNs)process information in the form of sparse spikes,creating the advantage of high energy efficiency for computer vision tasks.However,most studies have focused on simple classification tasks,and only a few researchers have applied SNNs to object detection in natural images.In this study,we consider the parsimonious nature of biological brains and propose a fast ANN-to-SNN conversion method for remote sensing image detection.We establish a fast sparse model for pulse sequence perception based on group sparse features and conduct transform-domain sparse resampling of the original images to enable fast perception of image features and encoded pulse sequences.In addition,to meet accuracy requirements in relevant remote sensing scenarios,we theoretically analyze the transformation error and propose channel self-decaying weighted normalization(CSWN)to eliminate neuron overactivation.We propose S3Det,a remote sensing image object detection model.Our experiments,based on a large publicly available remote sensing dataset,show that S3Det achieves an accuracy performance similar to that of the ANN.Meanwhile,our transformed network is only 24.32%as sparse as the benchmark and consumes only 1.46 W,which is 1/122 of the original algorithm’s power consumption.展开更多
基金Supported by CUP(J53C22003010006,J43C24000230007)ICREA2019.
文摘Spiking neural networks(SNNs)represent a biologically-inspired computational framework that bridges neuroscience and artificial intelligence,offering unique advantages in temporal data processing,energy efficiency,and real-time decision-making.This paper explores the evolution of SNN technologies,emphasizing their integration with advanced learning mechanisms such as spike-timing-dependent plasticity(STDP)and hybridization with deep learning architectures.Leveraging memristors as nanoscale synaptic devices,we demonstrate significant enhancements in energy efficiency,adaptability,and scalability,addressing key challenges in neuromorphic computing.Through phase portraits and nonlinear dynamics analysis,we validate the system’s stability and robustness under diverse workloads.These advancements position SNNs as a transformative technology for applications in robotics,IoT,and adaptive low-power AI systems,paving the way for future innovations in neuromorphic hardware and hybrid learning paradigms.
文摘Spiking neural networks(SNN)represent a paradigm shift toward discrete,event-driven neural computation that mirrors biological brain mechanisms.This survey systematically examines current SNN research,focusing on training methodologies,hardware implementations,and practical applications.We analyze four major training paradigms:ANN-to-SNN conversion,direct gradient-based training,spike-timing-dependent plasticity(STDP),and hybrid approaches.Our review encompasses major specialized hardware platforms:Intel Loihi,IBM TrueNorth,SpiNNaker,and BrainScaleS,analyzing their capabilities and constraints.We survey applications spanning computer vision,robotics,edge computing,and brain-computer interfaces,identifying where SNN provide compelling advantages.Our comparative analysis reveals SNN offer significant energy efficiency improvements(1000-10000×reduction)and natural temporal processing,while facing challenges in scalability and training complexity.We identify critical research directions including improved gradient estimation,standardized benchmarking protocols,and hardware-software co-design approaches.This survey provides researchers and practitioners with a comprehensive understanding of current SNN capabilities,limitations,and future prospects.
基金supported in part by the Basic Science Research Program through the National Research Foundation of Korea(NRF),funded by the Ministry of Education(NRF-2021R1A6A1A03039493).
文摘Spiking Neural Network(SNN)inspired by the biological triggering mechanism of neurons to provide a novel solution for plant disease detection,offering enhanced performance and efficiency in contrast to Artificial Neural Networks(ANN).Unlike conventional ANNs,which process static images without fully capturing the inherent temporal dynamics,our approach represents the first implementation of SNNs tailored explicitly for agricultural disease classification,integrating an encoding method to convert static RGB plant images into temporally encoded spike trains.Additionally,while Bernoulli trials and standard deep learning architectures likeConvolutionalNeuralNetworks(CNNs)and Fully Connected Neural Networks(FCNNs)have been used extensively,our work is the first to integrate these trials within an SNN framework specifically for agricultural applications.This integration not only refines spike regulation and reduces computational overhead by 30%but also delivers superior accuracy(93.4%)in plant disease classification,marking a significant advancement in precision agriculture that has not been previously explored.Our approach uniquely transforms static plant leaf images into time-dependent representations,leveraging SNNs’intrinsic temporal processing capabilities.This approach aligns with the inherent ability of SNNs to capture dynamic,timedependent patterns,making them more suitable for detecting disease activations in plants than conventional ANNs that treat inputs as static entities.Unlike prior works,our hybrid encoding scheme dynamically adapts to pixel intensity variations(via threshold),enabling robust feature extraction under diverse agricultural conditions.The dual-stage preprocessing customizes the SNN’s behavior in two ways:the encoding threshold is derived from pixel distributions in diseased regions,and Bernoulli trials selectively reduce redundant spikes to ensure energy efficiency on low-power devices.We used a comprehensive dataset of 87,000 RGB images of plant leaves,which included 38 distinct classes of healthy and unhealthy leaves.To train and evaluate three distinct neural network architectures,DeepSNN,SimpleCNN,and SimpleFCNN,the dataset was rigorously preprocessed,including stochastic rotation,horizontal flip,resizing,and normalization.Moreover,by integrating Bernoulli trials to regulate spike generation,ourmethod focuses on extracting themost relevant featureswhile reducingcomputational overhead.Using a comprehensivedatasetof87,000RGB images across 38 classes,we rigorously preprocessed the data and evaluated three architectures:DeepSNN,SimpleCNN,and SimpleFCNN.The results demonstrate that DeepSNN outperforms the other models,achieving superior accuracy,efficient feature extraction,and robust spike management,thereby establishing the potential of SNNs for real-time,energy-efficient agricultural applications.
基金financial supports from National Key Research and Development Program of China (2021YFB2801900,2021YFB2801901,2021YFB2801902,2021YFB2801904)National Natural Science Foundation of China (No.61974177)+1 种基金National Outstanding Youth Science Fund Project of National Natural Science Foundation of China (62022062)The Fundamental Research Funds for the Central Universities (QTZX23041).
文摘Neuromorphic photonic computing has emerged as a competitive computing paradigm to overcome the bottlenecks of the von-Neumann architecture.Linear weighting and nonlinear spike activation are two fundamental functions of a photonic spiking neural network(PSNN).However,they are separately implemented with different photonic materials and devices,hindering the large-scale integration of PSNN.Here,we propose,fabricate and experimentally demonstrate a photonic neuro-synaptic chip enabling the simultaneous implementation of linear weighting and nonlinear spike activation based on a distributed feedback(DFB)laser with a saturable absorber(DFB-SA).A prototypical system is experimentally constructed to demonstrate the parallel weighted function and nonlinear spike activation.Furthermore,a fourchannel DFB-SA laser array is fabricated for realizing matrix convolution of a spiking convolutional neural network,achieving a recognition accuracy of 87%for the MNIST dataset.The fabricated neuro-synaptic chip offers a fundamental building block to construct the large-scale integrated PSNN chip.
基金Supported by the National Natural Science Foundation of China (60904018, 61203040)the Natural Science Foundation of Fujian Province of China (2009J05147, 2011J01352)+1 种基金the Foundation for Distinguished Young Scholars of Higher Education of Fujian Province of China (JA10004)the Science Research Foundation of Huaqiao University (09BS617)
文摘For accelerating the supervised learning by the SpikeProp algorithm with the temporal coding paradigm in spiking neural networks (SNNs), three learning rate adaptation methods (heuristic rule, delta-delta rule, and delta-bar-delta rule), which are used to speed up training in artificial neural networks, are used to develop the training algorithms for feedforward SNN. The performance of these algorithms is investigated by four experiments: classical XOR (exclusive or) problem, Iris dataset, fault diagnosis in the Tennessee Eastman process, and Poisson trains of discrete spikes. The results demonstrate that all the three learning rate adaptation methods are able to speed up convergence of SNN compared with the original SpikeProp algorithm. Furthermore, if the adaptive learning rate is used in combination with the momentum term, the two modifications will balance each other in a beneficial way to accomplish rapid and steady convergence. In the three learning rate adaptation methods, delta-bar-delta rule performs the best. The delta-bar-delta method with momentum has the fastest convergence rate, the greatest stability of training process, and the maximum accuracy of network learning. The proposed algorithms in this paper are simple and efficient, and consequently valuable for practical applications of SNN.
基金supported by the National Natural Science Foundation of China(Nos.61974164,62074166,62004219,62004220,and 62104256).
文摘Artificial neural networks(ANNs)have led to landmark changes in many fields,but they still differ significantly fromthemechanisms of real biological neural networks and face problems such as high computing costs,excessive computing power,and so on.Spiking neural networks(SNNs)provide a new approach combined with brain-like science to improve the computational energy efficiency,computational architecture,and biological credibility of current deep learning applications.In the early stage of development,its poor performance hindered the application of SNNs in real-world scenarios.In recent years,SNNs have made great progress in computational performance and practicability compared with the earlier research results,and are continuously producing significant results.Although there are already many pieces of literature on SNNs,there is still a lack of comprehensive review on SNNs from the perspective of improving performance and practicality as well as incorporating the latest research results.Starting from this issue,this paper elaborates on SNNs along the complete usage process of SNNs including network construction,data processing,model training,development,and deployment,aiming to provide more comprehensive and practical guidance to promote the development of SNNs.Therefore,the connotation and development status of SNNcomputing is reviewed systematically and comprehensively from four aspects:composition structure,data set,learning algorithm,software/hardware development platform.Then the development characteristics of SNNs in intelligent computing are summarized,the current challenges of SNNs are discussed and the future development directions are also prospected.Our research shows that in the fields of machine learning and intelligent computing,SNNs have comparable network scale and performance to ANNs and the ability to challenge large datasets and a variety of tasks.The advantages of SNNs over ANNs in terms of energy efficiency and spatial-temporal data processing have been more fully exploited.And the development of programming and deployment tools has lowered the threshold for the use of SNNs.SNNs show a broad development prospect for brain-like computing.
基金supports from the National Key Research and Development Program of China (Nos.2021YFB2801900,2021YFB2801901,2021YFB2801902,2021YFB2801903,2021YFB2801904)the National Outstanding Youth Science Fund Project of National Natural Science Foundation of China (No.62022062)+1 种基金the National Natural Science Foundation of China (No.61974177)the Fundamental Research Funds for the Central Universities (No.QTZX23041).
文摘Spiking neural networks(SNNs)utilize brain-like spatiotemporal spike encoding for simulating brain functions.Photonic SNN offers an ultrahigh speed and power efficiency platform for implementing high-performance neuromorphic computing.Here,we proposed a multi-synaptic photonic SNN,combining the modified remote supervised learning with delayweight co-training to achieve pattern classification.The impact of multi-synaptic connections and the robustness of the network were investigated through numerical simulations.In addition,the collaborative computing of algorithm and hardware was demonstrated based on a fabricated integrated distributed feedback laser with a saturable absorber(DFB-SA),where 10 different noisy digital patterns were successfully classified.A functional photonic SNN that far exceeds the scale limit of hardware integration was achieved based on time-division multiplexing,demonstrating the capability of hardware-algorithm co-computation.
基金sponsored by Key‐Area Research and Development Program of Guangdong Province,No.2020B0404020005.
文摘Spiking Neural Network is known as the third-generation artificial neural network whose development has great potential.With the help of Spike Layer Error Reassignment in Time for error back-propagation,this work presents a new network called SpikeGoogle,which is implemented with GoogLeNet-like inception module.In this inception module,different convolution kernels and max-pooling layer are included to capture deep features across diverse scales.Experiment results on small NMNIST dataset verify the results of the authors’proposed SpikeGoogle,which outperforms the previous Spiking Convolutional Neural Network method by a large margin.
基金Project supported by the Natural Science Foundation of Chongqing(Grant No.cstc2021jcyj-msxmX0565)the Fundamental Research Funds for the Central Universities(Grant No.SWU021002)the Graduate Research Innovation Project of Chongqing(Grant No.CYS22242)。
文摘In recent years, spiking neural networks(SNNs) have received increasing attention of research in the field of artificial intelligence due to their high biological plausibility, low energy consumption, and abundant spatio-temporal information.However, the non-differential spike activity makes SNNs more difficult to train in supervised training. Most existing methods focusing on introducing an approximated derivative to replace it, while they are often based on static surrogate functions. In this paper, we propose a progressive surrogate gradient learning for backpropagation of SNNs, which is able to approximate the step function gradually and to reduce information loss. Furthermore, memristor cross arrays are used for speeding up calculation and reducing system energy consumption for their hardware advantage. The proposed algorithm is evaluated on both static and neuromorphic datasets using fully connected and convolutional network architecture, and the experimental results indicate that our approach has a high performance compared with previous research.
基金Project supported by the National Key Research and Development Program of China(Grant No.2018 YFB1306600)the National Natural Science Foundation of China(Grant Nos.62076207,62076208,and U20A20227)the Science and Technology Plan Program of Yubei District of Chongqing(Grant No.2021-17)。
文摘Spiking neural networks(SNNs) are widely used in many fields because they work closer to biological neurons.However,due to its computational complexity,many SNNs implementations are limited to computer programs.First,this paper proposes a multi-synaptic circuit(MSC) based on memristor,which realizes the multi-synapse connection between neurons and the multi-delay transmission of pulse signals.The synapse circuit participates in the calculation of the network while transmitting the pulse signal,and completes the complex calculations on the software with hardware.Secondly,a new spiking neuron circuit based on the leaky integrate-and-fire(LIF) model is designed in this paper.The amplitude and width of the pulse emitted by the spiking neuron circuit can be adjusted as required.The combination of spiking neuron circuit and MSC forms the multi-synaptic spiking neuron(MSSN).The MSSN was simulated in PSPICE and the expected result was obtained,which verified the feasibility of the circuit.Finally,a small SNN was designed based on the mathematical model of MSSN.After the SNN is trained and optimized,it obtains a good accuracy in the classification of the IRIS-dataset,which verifies the practicability of the design in the network.
基金This work is supported by the Ministry of Education Malaysia and Universiti Teknologi Malaysia through Research University Grant Scheme(Q.J130000.2651.16J63).
文摘Adolescent Idiopathic Scoliosis(AIS)is a deformity of the spine that affects teenagers.The current method for detecting AIS is based on radiographic images which may increase the risk of cancer growth due to radiation.Photogrammetry is another alternative used to identify AIS by distinguishing the curves of the spine from the surface of a human’s back.Currently,detecting the curve of the spine is manually performed,making it a time-consuming task.To overcome this issue,it is crucial to develop a better model that automatically detects the curve of the spine and classify the types of AIS.This research proposes a new integration of ESNN and Feature Extraction(FE)methods and explores the architecture of ESNN for the AIS classification model.This research identifies the optimal Feature Extraction(FE)methods to reduce computational complexity.The ability of ESNN to provide a fast result with a simplicity and performance capability makes this model suitable to be implemented in a clinical setting where a quick result is crucial.A comparison between the conventional classifier(Support Vector Machine(SVM),Multi-layer Perceptron(MLP)and Random Forest(RF))with the proposed AIS model also be performed on a dataset collected by an orthopedic expert from Hospital Universiti Kebangsaan Malaysia(HUKM).This dataset consists of various photogrammetry images of the human back with different types ofMalaysian AIS patients to solve the scoliosis problem.The process begins by pre-processing the images which includes resizing and converting the captured pictures to gray-scale images.This is then followed by feature extraction,normalization,and classification.The experimental results indicate that the integration of LBP and ESNN achieves higher accuracy compared to the performance of multiple baseline state-of-the-art Machine Learning for AIS classification.This demonstrates the capability of ESNN in classifying the types of AIS based on photogrammetry images.
文摘Medical image classification becomes a vital part of the design of computer aided diagnosis(CAD)models.The conventional CAD models are majorly dependent upon the shapes,colors,and/or textures that are problem oriented and exhibited complementary in medical images.The recently developed deep learning(DL)approaches pave an efficient method of constructing dedicated models for classification problems.But the maximum resolution of medical images and small datasets,DL models are facing the issues of increased computation cost.In this aspect,this paper presents a deep convolutional neural network with hierarchical spiking neural network(DCNN-HSNN)for medical image classification.The proposed DCNN-HSNN technique aims to detect and classify the existence of diseases using medical images.In addition,region growing segmentation technique is involved to determine the infected regions in the medical image.Moreover,NADAM optimizer with DCNN based Capsule Network(CapsNet)approach is used for feature extraction and derived a collection of feature vectors.Furthermore,the shark smell optimization algorithm(SSA)based HSNN approach is utilized for classification process.In order to validate the better performance of the DCNN-HSNN technique,a wide range of simulations take place against HIS2828 and ISIC2017 datasets.The experimental results highlighted the effectiveness of the DCNN-HSNN technique over the recent techniques interms of different measures.Please type your abstract here.
基金supported by the National Natural Science Foundation of China(62236002,62303009,62206001,52305001,62102387,62206005)the University Synergy Innovation Program of Anhui Province(GXXT-2022-041)the China Postdoctoral Science Foundation(2023M740013)
文摘Deep reinforcement learning(DRL)achieves success through the representational capabilities of deep neural networks(DNNs).Compared to DNNs,spiking neural networks(SNNs),known for their binary spike information processing,exhibit more biological characteristics.However,the challenge of using SNNs to simulate more biologically characteristic neuronal dynamics to optimize decision-making tasks remains,directly related to the information integration and transmission in SNNs.Inspired by the advanced computational power of dendrites in biological neurons,we propose a multi-dendrite spiking neuron(MDSN)model based on Multi-compartment spiking neurons(MCN),expanding dendrite types from two to multiple and deriving the analytical solution of somatic membrane potential.We apply the MDSN to deep distributional reinforcement learning to enhance its performance in executing complex decisionmaking tasks.The proposed model can effectively and adaptively integrate and transmit meaningful information from different sources.Our model uses a bioinspired event-enhanced dendrite structure to emphasize features.Meanwhile,by utilizing dynamic membrane potential thresholds,it adaptively maintains the homeostasis of MDSN.Extensive experiments on Atari games show that the proposed model outperforms some state-of-the-art spiking distributional RL models by a significant margin.
基金founded by the European Space Agency(contract number:4000135881/21/NL/GLC/my)in the framework of the Ariadna research program.
文摘While the exponential growth of the space sector and new operative concepts ask for higher spacecraft autonomy,the development of AI-assisted space systems was so far hindered by the low availability of power and energy typical of space applications.In this context,Spiking Neural Networks(SNN)are highly attractive because of their theoretically superior energy efficiency due to their inherently sparse activity induced by neurons communicating by means of binary spikes.Nevertheless,the ability of SNN to reach such efficiency on real world tasks is still to be demonstrated in practice.To evaluate the feasibility of utilizing SNN onboard spacecraft,this work presents a numerical analysis and comparison of different SNN techniques applied to scene classification for the EuroSAT dataset.Such tasks are of primary importance for space applications and constitute a valuable test case given the abundance of competitive methods available to establish a benchmark.Particular emphasis is placed on models based on temporal coding,where crucial information is encoded in the timing of neuron spikes.These models promise even greater efficiency of resulting networks,as they maximize the sparsity properties inherent in SNN.A reliable metric capable of comparing different architectures in a hardware-agnostic way is developed to establish a clear theoretical dependence between architecture parameters and the energy consumption that can be expected onboard the spacecraft.The potential of this novel method and its flexibility to describe specific hardware platforms is demonstrated by its application to predicting the energy consumption of a BrainChip Akida AKD1000 neuromorphic processor.
基金supports from National Natural Science Foundation of China(62171087,62475036).
文摘Photonic hardware implementation of spiking neural networks,regarded as a viable potential paradigm for ultra-high speed and energy efficiency computing,leverages spatiotemporal spike encoding and event-driven dynamics to simulate brain-like parallel information processing.Silicon-based microring resonators(MRRs)offer a power efficiency and ultrahigh flexibility scheme to mimic biological neuron,however,their substantial potential for integrated neuromorphic systems remains limited by insufficient exploration of MRR-based spiking digital and analog computation.Here,an all-optical neural dynamics framework,encompassing both excitatory and inhibitory behaviors based on multi-wavelength auxiliary and competition mechanism in an MRR,is proposed numerically.Leveraging multi-wavelength resonance characteristics and wavelength division multiplexing(WDM)technology,a single MRR implements the five fundamental optical digital logic gates:AND,OR,NOT,XNOR and XOR.Besides,the cascading capabilities of MRR-based spiking neurons are demonstrated through multi-level digital logic gates including NAND,NOR,4-input AND,8-input AND,and a full adder,emphasizing their promise for large-scale digital logic networks.Furthermore,an exemplary binary convolution has been achieved by utilizing the proposed MRR-based digital logic operation,illustrating the potential of all-optical binary convolution to compute image gradient magnitudes for edge detection.Such passive photonic neurons and networks promise access to the high transmission speed and low power consumption inherent to optical systems,thus enabling direct hardware-algorithm co-computation and accelerating artificial intelligence.
基金supported by the National Natural Science Foundation of China(62471190)the Natural Science Foundation of Hubei Province,China(2022CFA031)。
文摘Diffusive threshold switching(TS)memristors have emerged as a promising candidate for artificial neurons,effectively replicating neuronal functions and enabling spiking neural networks(SNNs)to emulate the low-power processing of biological brains.In this study,we present an artificial neuron based on a Pt/Ag/ZnO/Pt volatile memristor,which exhibits exceptional TS characteristics,including electro-forming-free operation,low voltage requirements(<0.2 V),high stability(2.25%variation over 1024 cycles),a high on/off ratio(106),and inherent self-compliance.These Pt/Ag/ZnO/Pt diffusive memristors are employed to simultaneously emulate oscillation neurons and leaky integrate-and-fire(LIF)neurons,enabling precise modulation of oscillation and firing frequencies through pulse parameters while maintaining low energy consumption(1.442 nJ per spike).We further integrate the oscillation and LIF neurons as input and output neurons,respectively,in a two-layer SNN,achieving a high classification accuracy of 89.17%on MNIST-based voltage images.This work underscores the potential of ZnO diffusive memristors in emulating stable artificial neurons and highlights their promise for advanced neuromorphic computing applications using SNNs.
基金supported by the Key-Area Research and Development Program of Guangdong Province(Grants No.2021B0909060002)National Natural Science Foundation of China(Grants No.62204219,62204140)Major Program of Natural Science Foundation of Zhejiang Province(Grants No.LDT23F0401).
文摘Spike-based neural networks,which use spikes or action potentialsto represent information,have gained a lot of attention because of their high energyefficiency and low power consumption.To fully leverage its advantages,convertingthe external analog signals to spikes is an essential prerequisite.Conventionalapproaches including analog-to-digital converters or ring oscillators,and sensorssuffer from high power and area costs.Recent efforts are devoted to constructingartificial sensory neurons based on emerging devices inspired by the biologicalsensory system.They can simultaneously perform sensing and spike conversion,overcoming the deficiencies of traditional sensory systems.This review summarizesand benchmarks the recent progress of artificial sensory neurons.It starts with thepresentation of various mechanisms of biological signal transduction,followed bythe systematic introduction of the emerging devices employed for artificial sensoryneurons.Furthermore,the implementations with different perceptual capabilitiesare briefly outlined and the key metrics and potential applications are also provided.Finally,we highlight the challenges and perspectives for the future development of artificial sensory neurons.
基金supported by the National Natural Science Foundation of China(Grant Nos.62376261,U21A20487)the Natural Science Foundation of Guangdong Province(Grant Nos.2024A1515011754,2023A1515011307,2022A1515140119).
文摘Significant progress has been made in brain-computer science and technology through applying spiking neural networks(SNNs)[1].More recently,due to its potential of processing complex spatio-temporal information,SNNs have been successfully applied in many fields,such as action recognition[2].There are two effective ways to design network models:converting artificial neural networks(ANNs)into SNNs and directly designing SNNs based on spike mechanisms.In the ANN-SNN method,the integrate-andfire(IF)neurons are used to replace the activation layer to convert ANNs into SNNs,which have some inherent drawbacks,such as inevitable accuracy loss,more delays and energy consumption.Although existing direct training strategies have shown outstanding performance in image classification tasks,SNNs face significant difficulties in handling complex video understanding tasks.In light of the considerable success achieved by the ANNs employed in the field of human action recognition,more researchers have recently focused their attention on using SNNs for action recognition.In order to design more efficient SNNs,some researchers have proposed a series of effective training and feature learning mechanisms in residual network,e.g.,Hu et al.
文摘The rapid advancement of deep learning and the emergence of largescale neural models,such as bidirectional encoder representations from transformers(BERT),generative pre-trained transformer(GPT),and large language model Meta AI(LLaMa),have brought significant computational and energy challenges.Neuromorphic computing presents a biologically inspired approach to addressing these issues,leveraging event-driven processing and in-memory computation for enhanced energy efficiency.This survey explores the intersection of neuromorphic computing and large-scale deep learning models,focusing on neuromorphic models,learning methods,and hardware.We highlight transferable techniques from deep learning to neuromorphic computing and examine the memoryrelated scalability limitations of current neuromorphic systems.Furthermore,we identify potential directions to enable neuromorphic systems to meet the growing demands of modern AI workloads.
基金Project supported by the National Key Research and Development Program of China(No.2022YFB4500900)。
文摘Artificial neural networks(ANNs)have made great strides in the field of remote sensing image object detection.However,low detection efficiency and high power consumption have always been significant bottlenecks in remote sensing.Spiking neural networks(SNNs)process information in the form of sparse spikes,creating the advantage of high energy efficiency for computer vision tasks.However,most studies have focused on simple classification tasks,and only a few researchers have applied SNNs to object detection in natural images.In this study,we consider the parsimonious nature of biological brains and propose a fast ANN-to-SNN conversion method for remote sensing image detection.We establish a fast sparse model for pulse sequence perception based on group sparse features and conduct transform-domain sparse resampling of the original images to enable fast perception of image features and encoded pulse sequences.In addition,to meet accuracy requirements in relevant remote sensing scenarios,we theoretically analyze the transformation error and propose channel self-decaying weighted normalization(CSWN)to eliminate neuron overactivation.We propose S3Det,a remote sensing image object detection model.Our experiments,based on a large publicly available remote sensing dataset,show that S3Det achieves an accuracy performance similar to that of the ANN.Meanwhile,our transformed network is only 24.32%as sparse as the benchmark and consumes only 1.46 W,which is 1/122 of the original algorithm’s power consumption.
