浏览全部资源
扫码关注微信
Received：23 May 2022，
Revised：2022-9-14，
Accepted：21 September 2022，
Published：16 February 2023
移动端阅览
相较于第1代和第2代神经网络，第3代神经网络的脉冲神经网络是一种更加接近于生物神经网络的模型，因此更具有生物可解释性和低功耗性。基于脉冲神经元模型，脉冲神经网络可以通过脉冲信号的形式模拟生物信号在神经网络中的传播，通过脉冲神经元的膜电位变化来发放脉冲序列，脉冲序列通过时空联合表达不仅传递了空间信息还传递了时间信息。当前面向模式识别任务的脉冲神经网络模型性能还不及深度学习，其中一个重要原因在于脉冲神经网络的学习方法不成熟，深度学习中神经网络的人工神经元是基于实数形式的输出，这使得其可以使用全局性的反向传播算法对深度神经网络的参数进行训练，脉冲序列是二值性的离散输出，这直接导致对脉冲神经网络的训练存在一定困难，如何对脉冲神经网络进行高效训练是一个具有挑战的研究问题。本文首先总结了脉冲神经网络研究领域中的相关学习算法，然后对其中主要的方法：直接监督学习、无监督学习的算法以及ANN2SNN的转换算法进行分析介绍，并对其中代表性的工作进行对比分析，最后基于对当前主流方法的总结，对未来更高效、更仿生的脉冲神经网络参数学习方法进行展望。
To understand the structure of human brain further
Wolfgang Mass summarizes that the structures
training methods and some other crucial parts of spiking neural networks (SNNs) systematically
which are known as the third-generation of artificial neural networks. There are hundreds of millions of neurons and synaptic structures in related to human brain
but the requirement of energy is quite small. The SNNs has its advantages of biological interpretability and lower power consumption in comparison with the first and the second generation artificial neural networks (ANNs). Its neurons simulate the internal dynamics of biological neurons
and the weight-balanced simulates the construction
enhancement and inhibition rules of biological synapses. The SNN is mainly composed of such commonly-used spiking neuron models in relevant to Hodgkin Huxley (HH) model
leaky integrate-and-fire (LIF) model
and spiking response (SRM) model. The difference of ion concentration-inner and the biological neuron-outer can activate the potential of the cell membrane. To improve an action potential
channel-based ions move in and out of the neuron membrane in the neuron membrane when a neuron is stimulated. Spiking neuron model is a mathematical model that simulates the action potential process of biological neuron. A spiking neuron receives neurons-derived spiking stimulation in the upper layer. It will fire spikes to outreach a spiking train. The SNNs is focused on the transmission from spiking trains to information-targeted
which can simulate the propagation of biological signals in the biological neural network. The spiking trains can convey spatiotemporal information. However
current performance of SNNs-based pattern recognition tasks is still challenged to its immature deep learning methods. The artificial neurons of the neural network are based on the output in the form of real numbers
which makes it possible to use the global back-propagation algorithm to train the parameters of the deep neural network. But the spiking train is a sort of binary discrete output
which is still a challenging issue for SNN-based training. First
to clarify its current situation
our summary is focused on recent SNN-based learning algorithms. Then
to analyze pros and cons of popular works
the three main algorithms are introduced: 1) supervised learning
2) unsupervised learning
and 3) ANN-SNN conversion. The unsupervised learning algorithm is mainly based on the mechanism of spike timing dependent plasticity (STDP). The biological synapses-interconnected is enhanced or inhibited according to the relative timing of the firing of presynaptic neurons and postsynaptic neurons. Unsupervised learning methods have stronger biological interpretability
which can use the local optimization method to balance the synaptic weights
but this method is challenged for its complicated and large-scale network structures. Therefore
drawing on the advantages of ANN's easy calculation
some supervised algorithms have emerged like gradient-based training method and ANN2SNN method. The gradient-based learning algorithm is mainly concerned of the training idea of back-propagation (BP)
which can balance the weight in terms of the error between the output value and the target value. This challenge is to be resolved in accordance with non-differentiable nature of discrete spikes. More methods of the BP-error have been proposed like gradient surrogate method. This gradient-based training method is focused on leveraging the training advantages of ANN and SNN. The training of SNN is interpretable biologically and easy to be computed. The ANN2SNN method can be used to convert the ANN weights-trained to SNN. This method can be used to realize the continuous activation values in the ANN into spiking trains. To reduce the conversion loss of ANN and SNN
this method is fine-tuned and converted according to neuron dynamics. This training method feature is indirect that it can apply SNN to complex network structures. The method of weight transfer can avoid direct training of SNN
which can apply SNN to complicated network structures. The ANN has been widely used in the field of image recognition. To extract more image features
ANN can be mainly used for consistent functions. SNN has its features of interpretability-biological and power consumption-lower
which can show its high performance in image recognition tasks. Finally
future SNNs-bionic learning methods are predicted in terms of some popular domain methods.
Bohte S M, Kok J N and La Poutré H. 2002. Error-backpropagation in temporally encoded networks of spiking neurons. Neurocomputing, 48(1-4): 17-37 [DOI: 10.1016/S0925-2312(01)00658-0]
Bu T, Ding J H, Yu Z F and Huang T J. 2022. Optimized potential initialization for low-latency spiking neural networks [EB/OL]. [2022-05-08] . https://arxiv.org/pdf/2202.01440.pdf https://arxiv.org/pdf/2202.01440.pdf
Cao Y Q, Chen Y and Khosla D. 2015. Spiking deep convolutional neural networks for energy-efficient object recognition. International Journal of Computer Vision, 113(1): 54-66 [DOI: 10.1007/s11263-014-0788-3]
Comsa I M, Potempa K, Versari L, Fischbacher T, Gesmundo A and Alakuijala J. 2020. Temporal coding in spiking neural networks with alpha synaptic function//Proceedings of ICASSP 2020-2020 IEEE International Conference on Acoustics. Barcelona, Spain: IEEE: 8529-8533[ DOI: 10.1109/ICASSP40776.2020.9053856 http://dx.doi.org/10.1109/ICASSP40776.2020.9053856 ]
Deng S K and Gu S. 2021. Optimal conversion of conventional artificial neural networks to spiking neural networks[EB/OL]. [2022-05-08] . https://arxiv.org/pdf/2103.00476.pdf https://arxiv.org/pdf/2103.00476.pdf
Deng S K, Li Y H, Zhang S H and Gu S. 2022. Temporal efficient training of spiking neural network via gradient re-weighting[EB/OL]. [2022-05-08] . https://arxiv.org/pdf/2202.11946.pdf https://arxiv.org/pdf/2202.11946.pdf
Diehl P U and Cook M. 2015. Unsupervised learning of digit recognition using spike-timing-dependent plasticity. Frontiers in Computational Neuroscience, 9: #99 [DOI: 10.3389/fncom.2015.00099]
Diehl P U, Neil D, Binas J, Cook M, Liu S C and Pfeiffer M. 2015. Fast-classifying, high-accuracy spiking deep networks through weight and threshold balancing//Proceedings of 2015 International Joint Conference on Neural Networks (IJCNN). Killarney, Ireland: IEEE: 1-8 [ DOI: 10.1109/IJCNN.2015.7280696 http://dx.doi.org/10.1109/IJCNN.2015.7280696 ]
Ding J H, Yu Z F, Tian Y H and Huang T J. 2021. Optimal ANN-SNN conversion for fast and accurate inference in deep spiking neural networks//Proceedings of the 30th International Joint Conference on Artificial Intelligence. Montreal, Canada: AAAI Press: 2328-2336 [ DOI: 10.24963/ijcai.2021/321 http://dx.doi.org/10.24963/ijcai.2021/321 ]
Fang W, Yu Z F, Chen Y Q, Huang T J, Masquelier T and Tian Y H. 2022. Deep residual learning in spiking neural networks[EB/OL]. [2022-01-22] . https://arxiv.org/pdf/2102.04159v6.pdf https://arxiv.org/pdf/2102.04159v6.pdf
Fang W, Yu Z F, Chen Y Q, Masquelier T, Huang T J and TianY H. 2021. Incorporating learnable membrane time constant to enhance learning of spiking neural networks//Proceedings of 2021 IEEE/CVF International Conference on Computer Vision. Montreal, Canada: IEEE: 2641-2651 [ DOI: 10.1109/ICCV48922.2021.00266 http://dx.doi.org/10.1109/ICCV48922.2021.00266 ]
Florian R V. 2012. The chronotron: a neuron that learns to fire temporally precise spike patterns. PLoS ONE, 7(8): #e40233 [DOI: 10.1371/journal.pone.0040233]
Gerstner W, Kistler W M, Naud R and Paninski L. 2014. NeuronalDynamics: from Single Neurons to Networks and Models of Cognition. Cambridge: Cambridge University Press
Guo S Q, Yu Z F, Deng F, Hu X L and Chen F. 2019. Hierarchical Bayesian inference and learning in spiking neural networks. IEEE Transactions on Cybernetics, 49(1): 133-145 [DOI: 10.1109/TCYB.2017.2768554]
Gütig R and Sompolinsky H. 2006. The tempotron: a neuron that learns spike timing-based decisions. Nature Neuroscience, 9(3): 420-428 [DOI: 10.1038/nn1643]
Han B and Roy K. 2020. Deep spiking neural network: energy efficiency through time based coding//Proceedings of the 16th European Conference on Computer Vision. Glasgow, UK: Springer: 388-404 [ DOI: 10.1007/978-3-030-58607-2_23 http://dx.doi.org/10.1007/978-3-030-58607-2_23 ]
Han B, Srinivasan G and Roy K. 2020. RMP-SNN: residual membrane potential neuron for enabling deeper high-accuracy and low-latency spiking neural network//Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, USA: IEEE: 13555-13564 [ DOI: 10.1109/CVPR42600.2020.01357 http://dx.doi.org/10.1109/CVPR42600.2020.01357 ]
Hebbian D O. 1949. The Organization of Behavior. A Neuropsychological Theory. New York: John Wiley&Sons
Hodgkin A L and Huxley A F. 1952. A quantitative description of membrane current and its application to conduction and excitation in nerve. The Journal of Physiology, 117(4): 500-544 [DOI: 10.1113/jphysiol.1952.sp004764]
Hu Y F, Li G Q, Wu Y J and Deng L. 2021. Spiking neural networks: a survey on recent advances and new directions. Control and Decision, 36(1): 1-26
胡一凡, 李国齐, 吴郁杰, 邓磊. 2021. 脉冲神经网络研究进展综述. 控制与决策, 36(1): 1-26 [DOI: 10.13195/j.kzyjc.2020.1006]
Hu Y F, Tang H J and Pan G. 2021a. Spiking deep residual networks. IEEE Transactions on Neural Networks and Learning Systems [DOI: 10.1109/TNNLS.2021.3119238]
Hu Y F, Wu Y J, Deng L and Li G Q. 2021b. Advancing residual learning towards powerful deep spiking neural networks[EB/OL]. [2021-12-23] . https://arxiv.org/pdf/2112.08954v1.pdf https://arxiv.org/pdf/2112.08954v1.pdf
Huang T J, Yu Z F, Li Y, Shi B X, Xiong R Q, Ma L and Wang W. 2022. Advances in spike vision. Journal of Image and Graphics, 27(6): 1823-1839
黄铁军, 余肇飞, 李源, 施柏鑫, 熊瑞勤, 马雷, 王威. 2022. 脉冲视觉研究进展. 中国图象图形学报, 27(6): 1823-1839 [DOI: 10.11834/jig.220175]
Kheradpisheh S R, Ganjtabesh M, Thorpe S J and Masquelier T. 2018. STDP-based spiking deep convolutional neural networks for object recognition. Neural Networks, 99: 56-67 [DOI: 10.1016/j.neunet.2017.12.005]
Kheradpisheh S R and Masquelier T. 2020. Temporal backpropagation for spiking neural networks with one spike per neuron. International Journal of Neural Systems, 30(6): #2050027 [DOI: 10.1142/S0129065720500276]
Kim J, Kim K and Kim J J. 2020. Unifying activation- and timing-based learning rules for spiking neural networks[EB/OL]. [2020-10-23] . https://arxiv.org/pdf/2006.02642v2.pdf https://arxiv.org/pdf/2006.02642v2.pdf
Lee C, Sarwar S S, Panda P, Srinivasan G and Roy K. 2020. Enabling spike-based backpropagation for training deep neural network architectures. Frontiers in Neuroscience, 14: #119 [DOI: 10.3389/fnins.2020.00119]
Li J N and Tian Y H. 2021. Recent advances in neuromorphic vision sensors: a survey. Chinese Journal of Computers, 44(6): 1258-1286
李家宁, 田永鸿. 2021. 神经形态视觉传感器的研究进展及应用综述. 计算机学报, 44(6): 1258-1286 [DOI: 10.11897/SP.J.1016.2021.01258]
Li Y H, Deng S K, Dong X, Gong R H and Gu S. 2021. A free lunch from ANN: towards efficient, accurate spiking neural networks calibration[EB/OL]. [2021-06-13] . https://arxiv.org/pdf/2106.06984.pdf https://arxiv.org/pdf/2106.06984.pdf
Liu F X, Zhao W B, Chen Y B, Wang Z W, Yang T and Jiang L. 2021. SSTDP: supervised spike timing dependent plasticity for efficient spiking neural network training. Frontiers in Neuroscience, 15: #756876 [DOI: 10.3389/fnins.2021.756876]
Liu Z Z, Chotibut T, Hillar C and Lin S W. 2020. Biologically plausible sequence learning with spiking neural networks. Proceedings of the AAAI Conference on Artificial Intelligence, 34(2): 1316-1323 [DOI: 10.1609/aaai.v34i02.5487]
Lyu M Y, Shao C P, Li H Y, Li J and Sun T F. 2021. A novel spiking neural network with the learning strategy of biomimetic structure//Proceedings of 2021 Asia-Pacific Conference on Communications Technology and Computer Science (ACCTCS). Shenyang, China: IEEE: 69-74 [ DOI: 10.1109/ACCTCS52002.2021.00022 http://dx.doi.org/10.1109/ACCTCS52002.2021.00022 ]
Mohemmed A, Schliebs S, Matsuda S and Kasabov N. 2012. Span: spike pattern association neuron for learning spatio-temporal spike patterns. International Journal of Neural Systems, 22(4): #1250012 [DOI: 10.1142/S0129065712500128]
Mozafari M, Ganjtabesh M, Nowzari-Dalini A, Thorpe S J and Masquelier T. 2019. Bio-inspired digit recognition using reward-modulated spike-timing-dependent plasticity in deep convolutional networks. Pattern Recognition, 94: 87-95 [DOI: 10.1016/j.patcog.2019.05.015]
Pfister J P and Gerstner W. 2006. Triplets of spikes in a model of spike timing-dependent plasticity. Journal of Neuroscience, 26(38): 9673-9682 [DOI: 10.1523/JNEUROSCI.1425-06.2006]
Ponulak F and Kasiński A. 2010. Supervised learning in spiking neural networks with ReSuMe: sequence learning, classification, and spike shifting. Neural Computation, 22(2): 467-510 [DOI: 10.1162/neco.2009.11-08-901]
Rathi N, Srinivasan G, Panda P and Roy K. 2020. Enabling deep spiking neural networks with hybrid conversion and spike timing dependent backpropagation[EB/OL]. [2022-05-08] . https://arxiv.org/pdf/2005.01807.pdf https://arxiv.org/pdf/2005.01807.pdf
Rueckauer B, Lungu I A, Hu Y H, Pfeiffer M and Liu S C. 2017. Conversion of continuous-valued deep networks to efficient event-driven networks for image classification. Frontiers in Neuroscience, 11: #682 [DOI: 10.3389/fnins.2017.00682]
Rumelhart D E, Hinton G E and Williams R J. 1986. Learning representations by back-propagating errors. Nature, 323(6088): 533-536 [DOI: 10.1038/323533a0]
Saunders D J, Siegelmann H T, Kozma R and Ruszinkó M. 2018. STDP learning of image patches with convolutional spiking neural networks//Proceedings of 2018 International Joint Conference on Neural Networks (IJCNN). Rio de Janeiro, Brazil: IEEE: 1-7 [ DOI: 10.1109/IJCNN.2018.8489684 http://dx.doi.org/10.1109/IJCNN.2018.8489684 ]
Sengupta A, Ye Y T, Wang R, Liu C A and Roy K. 2019. Going deeper in spiking neural networks: VGG and residual architectures. Frontiers in Neuroscience, 13: #95 [DOI: 10.3389/fnins.2019.00095]
Shahim-Aeen A and Karimi G. 2015. Triplet-based spike timing dependent plasticity (TSTDP) modeling using VHDL-AMS. Neurocomputing, 149: 1440-1444 [DOI: 10.1016/j.neucom.2014.08.050]
Shen J R, Zhao Y, Liu J K and Wang Y M. 2021. HybridSNN: combining bio-machine strengths by boosting adaptive spiking neural networks. IEEE Transactions on Neural Networks and Learning Systems [DOI: 10.1109/TNNLS.2021.3131356]
Shrestha S B and OrchardG. 2018. Slayer: spike layer error reassignment in time[EB/OL]. [2022-05-08] . https://arxiv.org/pdf/1810.08646.pdf https://arxiv.org/pdf/1810.08646.pdf
Song S, Miller K D and Abbott L F. 2000. Competitive Hebbian learning through spike-timing-dependent synaptic plasticity. Nature Neuroscience, 3(9): 919-926 [DOI: 10.1038/78829]
Stöckl C and Maass W.2021. Optimized spiking neurons can classify images with high accuracy through temporal coding with two spikes. Nature Machine Intelligence, 3(3): 230-238 [DOI: 10.1038/s42256-021-00311-4]
Taherkhani A, Belatreche A, Li Y H and Maguire L P. 2015. DL-ReSuMe: a delay learning-based remote supervised method for spiking neurons. IEEE transactions on neural networks and learning systems, 26(12): 3137-3149 [DOI: 10.1109/TNNLS.2015.2404938]
Tavanaei A and Maida A. 2019. BP-STDP: approximating backpropagation using spike timing dependent plasticity. Neurocomputing, 330: 39-47 [DOI: 10.1016/j.neucom.2018.11.014]
Thiele J C, Bichler O and Dupret A. 2019. Spikegrad: an ANN-equivalent computation model for implementing backpropagation with spikes[EB/OL]. [2022-05-08] . https://arxiv.org/pdf/1906.00851.pdf https://arxiv.org/pdf/1906.00851.pdf
Wade J J, McDaid L J, Santos J A and Sayers H M. 2010. SWAT: a spiking neural network training algorithm for classification problems. IEEE Transactions on Neural Networks, 21(11): 1817-1830 [DOI: 10.1109/TNN.2010.2074212]
Wu H, Zhang Y Y, Weng W M, Zhang Y T, Xiong Z W, Zha Z J, Sun X Y and Wu F. 2021. Training spiking neural networks with accumulated spiking flow. Proceedings of the AAAI Conference on Artificial Intelligence, 35(12): 10320-10328 [DOI: 10.1609/aaai.v35i12.17236]
Wu Y J, Deng L, Li G Q, Zhu J and Shi L P. 2018. Spatio-temporal backpropagation for training high-performance spiking neural networks. Frontiers in Neuroscience, 12: #331 [DOI: 10.3389/fnins.2018.00331]
Wu Y J, Deng L, Li G Q, Zhu J, Xie Y and Shi L P. 2019. Direct training for spiking neural networks: faster, larger, better. Proceedings of the AAAI Conference on Artificial Intelligence, 33(1): 1311-1318 [DOI: 10.1609/aaai.v33i01.33011311]
Xu Q, Qi Y, Yu H, Shen J R, Tang H J and Pan G. 2018. CSNN: an augmented spiking based framework with perceptron-inception//Proceedings of the 27th International Joint Conference on Artificial Intelligence. Stockholm, Sweden: AAAI Press: 1646-1652 [ DOI: 10.24963/ijcai.2018/228 http://dx.doi.org/10.24963/ijcai.2018/228 ]
Xu Y, Zeng X Q, Han L X and Yang J. 2013. A supervised multi-spike learning algorithm based on gradient descent for spiking neural networks. Neural Networks, 43: 99-113 [DOI: 10.1016/j.neunet.2013.02.003]
Yang X Y, Meng M Y, Xiao S L and Yu Z Y. 2021. SPA: stochastic probability adjustment for system balance of unsupervised SNNs//Proceedings of the 25th International Conference on Pattern Recognition (ICPR). Milan, Italy: IEEE: 6417-6424 [ DOI: 10.1109/ICPR48806.2021.9412266 http://dx.doi.org/10.1109/ICPR48806.2021.9412266 ]
Yu Q, Tang H J, Tan K C and Li H Z. 2013. Precise-spike-driven synaptic plasticity: learning hetero-association of spatiotemporal spike patterns. PLoS ONE, 8(11): #e78318 [DOI: 10.1371/journal.pone.0078318]
Zhang W R and Li P. 2021. Temporal spike sequence learning via backpropagation for deep spiking neural networks [EB/OL].[2022-05-08] . https://arxiv.org/pdf/2002.10085.pdf https://arxiv.org/pdf/2002.10085.pdf
Zheng H L, Wu Y J, Deng L, Hu Y F and Li G Q. 2021. Going deeper with directly-trained larger spiking neural networks. Proceedings of the AAAI Conference on Artificial Intelligence, 35(12): 11062-11070 [DOI: 10.1609/aaai.v35i12.17320]
相关文章
相关作者
相关机构
京公网安备11010802024621
