脉冲神经网络的监督学习算法研究综述

doi:10.3969/j.issn.0372-2112.2015.03.024

摘要/Abstract

摘要：

脉冲神经网络是进行复杂时空信息处理的有效工具,但由于其内在的不连续和非线性机制,构建高效的脉冲神经网络监督学习算法非常困难,同时也是该研究领域的重要问题.本文介绍了脉冲神经网络监督学习算法的基本框架,以及性能评价原则,包括脉冲序列学习能力、离线与在线处理性能、学习规则的局部特性和对神经网络结构的适用性.此外,对脉冲神经网络监督学习算法的梯度下降学习规则、突触可塑性学习规则和脉冲序列卷积学习规则进行了详细的讨论,通过对比分析指出现有算法存在的优缺点,并展望了该领域未来的研究方向.

关键词: 脉冲神经网络, 监督学习, 反向传播, 突触可塑性, 卷积

Abstract:

Spiking neural networks are shown to be suitable tools for the processing of spatio-temporal information.However, due to their intricately discontinuous and implicit nonlinear mechanisms, the formulation of efficient supervised learning algorithms for spiking neural networks is difficult, which is an important problem in the research area.In this paper, we introduce the general framework of supervised learning algorithms for spiking neural networks, and analyze their performance evaluations including spike trains learning ability, offline and online processing ability, the locality of learning mechanism and the applicability to network structure.Furthermore, we survey the advance of the research on supervised learning algorithms, which can be divided into three categories according to their differences:gradient descent rule, synaptic plasticity rule, and spike trains convolution rule.Finally, we discuss the advantages and disadvantages of these algorithms, and prospect the problems in current research and some future research directions in this area.

Key words: spiking neural network, supervised learning, backpropagation, synaptic plasticity, convolution

中图分类号:

TP183

蔺想红, 王向文, 张宁, 马慧芳. 脉冲神经网络的监督学习算法研究综述[J]. 电子学报, 2015, 43(3): 577-586.

LIN Xiang-hong, WANG Xiang-wen, ZHANG Ning, MA Hui-fang. Supervised Learning Algorithms for Spiking Neural Networks:A Review[J]. Acta Electronica Sinica, 2015, 43(3): 577-586.

参考文献

[1] Haykin S S.Neural Networks and Learning Machines[M].Upper Saddle River:Pearson Education, 2009.
[2] Izhikevich E M.Which model to use for cortical spiking neurons?[J].IEEE Transactions on Neural Networks, 2004, 15(5):1063-1070.
[3] Bohte S M.The evidence for neural information processing with precise spike-times:A survey[J].Natural Computing, 2004, 3(2):195-206.
[4] Ghosh-Dastidar S, Adeli H.Spiking neural networks[J].International Journal of Neural Systems, 2009, 19(4):295-308.
[5] Knudsen E I.Supervised learning in the brain[J].Journal of Neuroscience, 1994, 14(7):3985-3997.
[6] Kasiński A, Ponulak F.Comparison of supervised learning methods for spike time coding in spiking neural networks[J].International Journal of Applied Mathematics and Computer Science, 2006, 16(1):101-113.
[7] Quiroga R Q, Panzeri S.Principles of Neural Coding[M].Boca Raton, FL:CRC Press, 2013.
[8] 蔺想红, 张田文.分段线性脉冲神经元模型的动力学特性分析[J].电子学报, 2009, 37(6):1270-1276. Lin Xiang-hong, Zhang Tian-wen.Dynamical properties of piecewise linear spiking neuron model[J].Acta Electronica Sinica, 2009, 37(6):1270-1276.(in Chinese)
[9] Brette R, Rudolph M, Carnevale T, et al.Simulation of networks of spiking neurons:A review of tools and strategies[J].Journal of Computational Neuroscience, 2007, 23(3):349-398.
[10] Naud R, Gerhard F, Mensi S, et al.Improved similarity measures for small sets of spike trains[J].Neural Computation, 2011, 23(12):3016-3069.
[11] Wang J, Belatreche A, Maguire L, et al.Online versus offline learning for spiking neural networks:A review and new strategies[A].Proceedings of the 9th International Conference on Cybernetic Intelligent Systems[C].London, UK:IEEE, 2010.1-6.
[12] Kasabov N K.Neu Cube:A spiking neural network architecture for mapping, learning and understanding of spatio-temporal brain data[J].Neural Networks, 2014, 52:62-76.
[13] Oniz Y, Kaynak O.Variable-structure-systems based approach for online learning of spiking neural networks and its experimental evaluation[J].Journal of the Franklin Institute, 2014, 351(6):3269-3285.
[14] Rumelhart D E, Hinton G E, Williams R J.Learning representations by back-propagating errors[J].Nature, 1986, 323(9):533-536.
[15] Bohte S M, Kok J N, La Poutré J A.Error-backpropagation in temporally encoded networks of spiking neurons[J].Neurocomputing, 2002, 48(1-4):17-37.
[16] Gerstner W, Kistler W M.Spiking Neuron Models:Single Neurons, Populations, Plasticity[M].Cambridge:Cambridge University Press, 2002.
[17] Xin J, Embrechts M J.Supervised learning with spiking neuron networks[A].Proceedings of the International Joint Conference on Neural Networks[C].Washington DC:IEEE, 2001.1772-1777.
[18] Schrauwen B, Van Campenhout J.Extending SpikeProp[A].Proceedings of the International Joint Conference on Neural Networks[C].Budapest, Hungary:IEEE, 2004.471-475.
[19] Mc Kennoch S, Liu D, Bushnell L G.Fast modifications of the SpikeProp algorithm[A].Proceedings of the International Joint Conference on Neural Networks[C].Vancouver, Canada:IEEE, 2006.3970-3977.
[20] Mc Kennoch S, Voegtlin T, Bushnell L.Spike-timing error backpropagation in theta neuron networks[J].Neural Computation, 2009, 21(1):9-45.
[21] Fang H, Luo J, Wang F.Fast learning in spiking neural networks by learning rate adaptation[J].Chinese Journal of Chemical Engineering, 2012, 20(6):1219-1224.
[22] Yang W, Yang J, Wu W.A modified spiking neuron that involves derivative of the state function at firing time[J].Neural Processing Letters, 2012, 36(2):135-144.
[23] Booij O, Nguyen T H.A gradient descent rule for spiking neurons emitting multiple spikes[J].Information Processing Letters, 2005, 95(6):552-558.
[24] Booij O.Temporalpattern Classification Using Spiking Neural Networks[D].Amsterdam:University of Amsterdam, 2004.
[25] 方慧娟, 王永骥.多脉冲发放的Spiking神经网络[J].应用科学学报, 2008, 26(6):638-644. Fang Hui-juan, Wang Yong-ji.Spiking neural networks with neurons firing multiple spikes[J].Journal of Applied Sciences, 2008, 26(6):638-644.(in Chinese)
[26] Ghosh-Dastidar S, Adeli H.A new supervised learning algorithm for multiple spiking neural networks with application in epilepsy and seizure detection[J].Neural Networks, 2009, 22(10):1419-1431.
[27] Xu Y, Zeng X, Han L, et al.A supervised multi-spike learning algorithm based on gradient descent for spiking neural networks[J].Neural Networks, 2013, 43:99-113.
[28] 巩祖正.脉冲神经网络的多脉冲定时误差反向传播算法研究[D].甘肃兰州:西北师范大学, 2013. Gong Zu-zheng.Multi-spike Timing Error Back Propagation Algorithm in Spiking Neural Networks[D].Lanzhou, Gansu:Northwest Normal University, 2013.(in Chinese)
[29] Tiňo P, Mills A J S.Learning beyond finite memory in recurrent networks of spiking neurons[J].Neural Computation, 2006, 18(3):591-613.
[30] Gütig R, Sompolinsky H.The Tempotron:A neuron that learns spike timing-based decisions[J].Nature Neuroscience, 2006, 9(3):420-428.
[31] Florian R V.The Chronotron:A neuron that learns to fire temporally precise spike patterns[J].PLoS One, 2012, 7(8):e40233.
[32] Victor J D, Purpura K P.Metric-space analysis of spike trains:Theory, algorithms and application[J].Network:Computation in Neural Systems, 1997, 8(2):127-164.
[33] Xu Y, Zeng X, Zhong S.A new supervised learning algorithm for spiking neurons[J].Neural Computation, 2013, 25(6):1472-1511.
[34] Le Mouel C, Harris K D, Yger P.Supervised learning with decision margins in pools of spiking neurons[J].Journal of Computational Neuroscience, 2014, 37(2):333-344
[35] Hebb D O.The Organization of Behavior:A Neuropsychological Theory[M].New York:Wiley, 1949.
[36] Caporale N, Dan Y.Spike timing-dependent plasticity:A Hebbian learning rule[J].Annual Review of Neuroscience, 2008, 31(1):25-46.
[37] Ruf B, Schmitt M.Learning temporally encoded patterns in networks of spiking neurons[J].Neural Processing Letters, 1997, 5(1):9-18.
[38] Legenstein R, Naeger C, Maass W.What can a neuron learn with spike-timing-dependent plasticity?[J].Neural Computation, 2005, 17(11):2337-2382.
[39] Ponulak F, Kasinski A.Supervised learning in spiking neural networks with ReSuMe:Sequence learning, classification, and spike shifting[J].Neural Computation, 2010, 22(2):467-510.
[40] Ponulak F.Analysis of the ReSuMe learning process for spiking neural networks[J].International Journal of Applied Mathematics and Computer Science, 2008, 18(2):117-127.
[41] Glackin C, Maguire L, Mc Daid L, et al.Receptive field optimisation and supervision of a fuzzy spiking neural network[J].Neural Networks, 2011, 24(3):247-256.
[42] Hu J, Tang H, Tan K C, et al.A spike-timing-based integrated model for pattern recognition[J].Neural Computation, 2013, 25(2):450-472.
[43] Sporea I, Grüning A.Supervised learning in multilayer spiking neural networks[J].Neural Computation, 2013, 25(2):473-509.
[44] Wade J J, Mc Daid L J, Santos J A, et al.SWAT:A spiking neural network training algorithm for classification problems[J].IEEE Transactions on Neural Networks, 2010, 21(11):1817-1830.
[45] Gardner B, Grüning A.Learning temporally precise spiking patterns through reward modulated spike-timing-dependent plasticity[A].Proceedings of the 23rd International Conference on Artificial Neural Networks[C].Sofia, Bulgaria:Springer Berlin Heidelberg, 2013.256-263.
[46] Paugam-Moisy H, Martinez R, Bengio S.Delay learning and polychronization for reservoir computing[J].Neurocomputing, 2008, 71(7):1143-1158.
[47] Pfister J P, Toyoizumi T, Barber D, et al.Optimal spike-timing-dependent plasticity for precise action potential firing in supervised learning[J].Neural Computation, 2006, 18(6):1318-1348.
[48] Brea J, Senn W, Pfister J P.Matching recall and storage in sequence learning with spiking neural networks[J].The Journal of Neuroscience, 2013, 33(23):9565-9575.
[49] Carnell A, Richardson D.Linear algebra for times series of spikes[A].Proceedings of the 13th European Symposium on Artificial Neural Networks[C].Evere, Belgium:d-side, 2005.363-368.
[50] Mohemmed A, Schliebs S.SPAN:Spike pattern association neuron for learning spatio-temporal spike patterns[J].International Journal of Neural Systems, 2012, 22(4):1250012.
[51] Mohemmed A, Schliebs S, Matsuda S, Kasabov N.Training spiking neural networks to associate spatio-temporal input-output spike patterns[J].Neurocomputing, 2013, 107:3-10.
[52] Yu Q, Tang H, Tan K C, Li H.Precise-spike-driven synaptic plasticity:Learning hetero-association of spatiotemporal spike patterns[J].PLoS One, 2013, 8(11):e78318.
[53] Yu Q, Tang H, Tan K C, Yu H.A brain-inspired spiking neural network model with temporal encoding and learning[J].Neurocomputing, 2014, 138:3-13. [LL]
[54] Mohemmed A, Kasabov N.Incremental learning algorithm for spatio-temporal spike pattern classification[A].Proceedings of the International Joint Conference on Neural Networks[C].Brisbane, QLD:IEEE, 2012.1-6.
[55] Li C, Lu J, Wu C, et al.Bidirectional modification of presynaptic neuronal excitability accompanying spike timing-dependent synaptic plasticity[J].Neuron, 2004, 41(2):257-268.
[56] Wu W, Srivastava A.An information-geometric framework for statistical inferences in the neural spike train space[J].Journal of Computational Neuroscience, 2011, 31(3):725-748.
[57] Park I M, Seth S, Paiva A, et al.Kernel methods on spike train space for neuroscience:A tutorial[J].IEEE Signal Processing Magazine, 2013, 30(4):149-160.
[58] Selvaratnam K, Kuroe Y, Mori T.Learning methods of recurrent spiking neural networks-Transient and oscillatory spike trains[J].Transactions of the Institute of Systems, Control and Information Engineers, 2000, 44(3):95-104.
[59] Kuroe Y, Ueyama T.Learning methods of recurrent spiking neural networks based on adjoint equations approach[A].Proceedings of the International Joint Conference on Neural Networks[C].Barcelona, Spain:IEEE, 2010.1-8.
[60] Rostro-Gonzalez H, Cessac B, Viéville T.Parameter estimation in spiking neural networks:A reverse-engineering approach[J].Journal of Neural Engineering, 2012, 9(2):026024.