脉冲神经网络的监督学习算法研究综述

所需积分/C币:50 2018-01-30 20:34:52 667KB PDF
86
收藏 收藏
举报

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
2 3 h 4 Xin embrechts Spiker P op MeKennoch RPro Quick Prop plkcProp 14 ike rop Be dclta guven E VE VE E Spikeprop SRM 3.2.1 Boht Spikeprop ike re hosh astidar Adeli”6 sponse Modcl SRM Multi SpikeProp 21994-2017ChinaAcademicJournalElectronicPublishingHouse.Allrightsreservedhttp://www.cnki.net 580 2015 Florian Chrontron ⅩOR Fisher iris E Learning Multi-Spike Prop Learning Chrontron Earn Ing Victo urpur a 32 Tempotron Chrontron 3 E 3.3 26 4 RM 1~4、7 5、6、8 Tino Mills SpikeProp 6 Gutig Sompolinsky empotron SRM Integrate-and-F Integrate-and -F Tempotron I Bohte et al. I5 /Spike Prop 2MeKennoch et al. 19/RProp QuickProp 3BC guye 4 Ghosh - Dastidar Adeli 26/ Multi Spike Prop 5 Xu Yan et al tInn Mill 7Gutig Sompolinsky/Tempoto 8 Florian 31/ Chonotron E--earnin 21994-2017cHinaAcademicJOurnalElectronicPublishingHousc.alLrightsreservedhttp://www.cnki.net 581 STDP anti-STDP Remole Supervised Method Resume Hebb Resum Spike timing ependent Pl d ticity STDP D hebbian Hebb STDP Hebbian y STDP TDP ar s ti STDP Resume STDP 4.2.1 Hebbian u Me Ruf Schmitt Hebbian runing 4.2.3 TDH Wade bCM Bie tock-co T STDP SWAT Synaptic Weight Association Training Lewenstein STDE Hebbian Flo an Chrontron L-earning SWAT ig 7 sign 1:∑l,1-∑ Kw K orc ponulak kaminski 39 21994-2017cHinaAcademicJOurnalElectronicPublishingHousc.alLrightsreservedhttp://www.cnki.net 582 2015 STDP 4.3 STDP Paugam-Me sTDP Polvchronization Piste 6、8 -4、7、8 56 47 Kullback eibler 1 Ruf Schmitt 3 Florian 3/ Chonotron H-eurnin 4 Ponulak Kosinski 39/ReSuMe 5Sporea Ruining 6 Wade et al. 4/SWAT Paugam-moisy et Resume WAT Gruning ResuMe Carnell Richards 5.1 K L wo s t wiU,exp 6 21994-2017ChinaAcademicJournalElectronicPublishingHouse.Allrightsreservedhttp://www.cnki.net 583 E Gram Schmidt E/norm s. t norm s. t yIS t S. t ) SPAN PSD 5.2.2 SPAN Mohammed Schlicbs 5)51 SPAN Spike Pattern association Neuron Widrow-off CArnell Richard A s I dt SPAN Intcgratc-and - Firo eTe-/'T Heaviside Mohammed Schliebs 50 /sPAN Lte-t'I T exp 3 Yu et al. 5/PSD y 14- T已xp 15 5.2.3PSD Y SPAN Wid row-Hoff PSD Precise-Spike - Driven SPAN Spikcprop K I t一 16 K Widrow-Hoff PSD △,1=mS。t ∑∑Kt5-1H-1 SKI 18 PSD 21994-2017cHinaAcademicJOurnalElectronicPublishingHousc.alLrightsreservedhttp://www.cnki.net 584 2015 STDP Resume BP 1 Haykin SS Neural Netw orks and Learning M achines M U pper saddle River Pearson Education 2009 2 Izhikevich E M. Which model to use for cortical spiking neurons J. IEEE Transactions on Neural Netw orks 200415 l063-1070 3 Bohte S M. The evidence for neural information processing w ilh precise spike-imes A survey J. N:alural Com puling 200432195-206 4 Ghosh -Dastidar S Adeli H. Spiking neural netw orks J International Journal of Neural Systems 2009 19 4 295 Hilbert 308 5 Knudsen E l Supervised learning in the brain J. Journal SPAN” 6 Kasinski A Ponulak F Com parison of supervised learnin PSD 2 Inethods for spike Lime coding in spiking neural netw orks J. International Journal of Applied Mathematics and cience2006161101-113 7 Quiroga RQ Panzeri s Principles of Neural Coding M Boca raton fl crc press 2013 20093761270-1276 Lin Xiang iong Zhang Tian-wen. Dy narnical properties of piecew ise 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 net- works of spiking neurons A review of tools and strategies J. Journal of Computational N 2007233 349-398 10 Naud R Gerhard F Mensi S et al. Improved similarity for small sets of spike I. Neural 20l123123016-3069 11 Wang J Belatreche A Maguire L et al. Online versus off learning for spik in new strategies A. Proceedings of the th Internationa 21994-2017ChinaAcademicJournalElectronicPublishingHouse.Allrightsreservedhttp://www.cnki.net 585 Conference on Cy bernetic Intelligent Systems C. Lon- gorithm for multiple spiking neural networks w ith applica- lUn uk ieee 201o, 1-6 Lion in epilepsy and seizure detection J. Neural Nel 12 Kasabov n K. Neu Cube A spiking neural netw ork archi works200922101419-1431 tecture for mapping learning and understanding of spatio- 27 Xu Y Zeng X Han L ct al. A supervised multi-spike temporal brain data J. Neural Networks 2014 52 62 learning algorithm based on gradient descent for spiking neural networks J. Neural Nelw orks 2013 43 99 13 Oniz Y Kaynak 0. Variable structure -systems based proach for online learning of spiking neural networks and its experimental evaluation J. Journal of the franklin Ir D 2013 sTIlle201435163269-3285 Gong Z-yheng. Mulli-spike Timing Error Back Propaga- 14 Rumelhart D E Hinton G E Williams R J Learning rep tion Algorithm in Spiking Neural Netw orks D resentations by back propagating errors J. Nature 1986 Lanzhou gansu Northw est Normal University 2013. in 3239533-536. Chinese 15 Bohle SM Kok jn La poule JA. Frror-tackpropaga- 29 Tino P Mills a J S Learming beyond finite memory in re- tion in temporally enco ded netw orks of spiki sp g current netw orks of spiking neurons I. Neural Computa- J. Neurocomputing 2002 48 17-37 tion2006183591-613 16 Gerstner W Kistler W M. Spiking Neuron Models Single 30 Gutig R Sompolinsky H. The Tempotron A neuron that Neurons Populations Plaslicily M. Cambridge Cam- learns spike liming -based decisions J. Nalure Neuro bridge University Press 2002 science200693420-428. 17 Xin J Embrechts M J Supervised learning with spiking 31 Florian v. The Chronotron A neuron that learns to fire neuron netw orks A. Proceedings of the International temporally precise spike pa J. PLos One 2012 7 Joint Conference on Neural Netw orks C. Washington DC IEEE2001.1772-1777 32 Victor J D Purpura K P Metric-space analysis of spike 18 Schrauwen b Van Campenhout J Extending Spike Prop trains Theory algorithms and application J. Network A. Proceedings of the International Joint Conference on Computation in Neural Systems 19978 2 127-164 Neural Networks C. Budapest Hungary IEFF 2004 33 Xu Y Zeng X Zhong S. A new superv ised learning aIyo- 471-475 rithm for spiking neurons I. Neural Computation 2013 19 Mc Kennoch s Liu D Bushnell L G. fast modifications 2561472-15l1 of the SpikeProp algor A. Proceedings of the inter- 34 Le Mouel C Harris K d Yger P Supervised learning with nalional Joinl Conference on Neural Nelw orks C. vall- ecision margins in pools of spiking neurons J. Journal couver Canada IEEE 2006. 3970-3977 of Computational Neuroscience 2014 37 2 333-344 20 Mc Kennoch S Voegtlin ' T' Bushnell L Spike -timing error 35 Hebb DO. The Organization of Behavior A Neuropsy backpropagation in theta neuron networks J. n chological Theory M. New York Wiley 1949 Computation 2009 21 19-45 36 Caporale N Dan Y Spike timing -dependent plasticity A 1 Fang II Luo J Wang F Fast learning in spiking neural Ilebbian learning rule J. Annual Review of Neuro networks by learning rate adaptation J. Chinese Journa science20083l125-46. of Chemical Engineering 2012 20 6 1219-1224 37 Ruf B Schmitt M. Learning temporally encoded patterns 2 Yang w Yang J Wuw. a modified spiking neuron that in netw orks of spiking neurons J. Neural Processing involves derivative of the state function at firing time j Letters 1997 519-18 Neural Processing Letters 2012 36 2 135-1 38 Legcnstcin R Nacger C Maass W. What can a neuron 23 BooijO Nguyen T H a gradient descent rule for spiking learn with spike -timing-dependent plasticity J. Neural neurons emitting multiple spikes J. Information Process Computation200517112337-2382 2 7& Letters2005956552-558 39 Ponulak F Kasinski A Supervised learning in spiking Booij o. 'Tem poralpattern Classification Using Spiking neural networks with ResuMe sequence learning classi Neural Networks D. Amsterdam University of Amster- fication and spike shifting J. Neural Computation dam 2004 2010222467-510 25 40 Ponulak F. Analysis of the ReSuMe learning process for 2008266638-614. spiking neural networks J. International Journal of Ap- F Yons iking neural netw orks plied M athematics and Computer Science 2008 18 2 with neurons firing multiple spikes J. Journal of A pplie Sciences 2008 266 638-644. in Chinese 41 Glackin C Maguire L Mc Daid L et al. Receptive field 26 Ghosh -Dastidar S Adeli H. a new supervised learning al optimisation and supervision of a fuzzy spiking neural net 21994-2017cHinaAcademicJOurnalElectronicPublishingHousc.alLrightsreservedhttp://www.cnki.net 586 2015 work J. Neural Networks 2011 24 3 247-256 54 Mohemmed A Kasabov N Incremental learning algorithm 42 Hu J Tang H Tan K C el al. A spike-iming -) ased in Temporal spike pallern classification A. Pro grated model for pattern recognition J. Neural Computa ceedings of the International Joint Conference on Neural tion2013252450-472 Netw orks C. Brisbane QLD IEEE 2012. 1-6 43 Sporea I Gruning A. Supervised learning in multilayer 55 Li C Lu J WuC et al. Bidirectional modification of pres- spiking neural netw orks I. Neural Computation 2013 ynaptic neuronal excitability accompanying spike timing 252473-509 dependent sy naptic plasticity J. Ncuron 2004 41 2 44 Wade J J Mc Daid L J Santos J A et al. SWAT A spi- 257-268 kiny neural nel work Training algorithm for classification 56 Wu w Srivastava A. An informal ion-eomelrie: frame- problems IEEE Transactions on Neural Netw orks work for statistical inferences in the neural spike 20102l111817-1830. space J. Journal of Computational Neuroscience 2011 45 Gardner B Gruning A Learning temporally precise spi 313725-748 king patterns through rew ard modulated spike -timing-fle- 57 Park I M Seth S Paiva A et al. Kernel methods on spike pendent plasticity A. Proceedings of the 23 rd Interna train space for neuro science A tutorial J. IEEE Signal tional Conference on Artificial Neural Netw orks c. s Processing magazine 2013 30 4 149-160 fia Bulgaria Springer Berlin Heidelberg 2013. 256- 58 Selvaratnam K Kuroe Y Mori T Learning methods of re- currenl spiking neural nelw orks-TranIsienl and oscillatory 46 Paugam-Moisy H Martinez R Bengio S. Delay learning spike trains Transactions of the Institute of System and polychronization for reservoir computing J. Neuro Control and Inform ation engineers 2000 44 3 95 computing20087171143-1158 104 47 Pfister JP Toyoizumi t Barber D et al. Optimal spike 59 Kuroe Y Ueyama T Learning methods of recunent spi- timing dependent plasticity for precise action potential fir- king neural networks based on adjoint equations approach ing in supervised learning J. Neural Computation 2006 A. Proceedings of the International Joint Conference on 1861318-1348. Neural Nelworks C. Barcelona Spain IEEE. 2010.1 48 Brea j Senn w pfister JP. Matching recall and storage in 8 sequence learning w ith spiking neural netw orks J. The 60 Rostro-Gonzalez H Cessac B Vieville T. Parameter esti- Journal of Neuroscience 2013 33 23 9565-9575 mation in spiking neural netw orks A reverse-engineering 49 Carnell a richardson d. liuear algebra for limes series approach J. Journal of Neural Engineering 2012 9 2 of spikes A. Proceedings of the 13th European Symposi U26024 um on artificial Neural Netw orks C. Evere belgium d side2005.363-368 50 Mohemmed A Schliebs S. SPaN Spike pattern associa- 19761 tion neuron for learning spatio -temporal spike pattems 2009 J. International Journal of Neural Systems 2012 22 41250012 51 Mohemmed a schliebs s Matsuda s Kasaboy n. train- ing spiking neural networks to associate spatio temporal E-mail linxh nwnu. edu. cn input-output spike patterns J. Neurocomputing 2013 52 YuQ Tang H Tan K c Li h. Precise-spike -driven syn- 1993 aptic plasticity Learning hetero-association of spatiotem sOne2013811e78318 53 Yu Q Tang H Tan Kc Yu H. A brain-inspired spiking 2@163 tempora d le ing J. Neurocomputing 2014 138 3-13 21994-2017cHinaAcademicJOurnalElectronicPublishingHousc.alLrightsreservedhttp://www.cnki.net

...展开详情
立即下载 身份认证VIP会员低至7折
一个资源只可评论一次,评论内容不能少于5个字
您会向同学/朋友/同事推荐我们的CSDN下载吗?
谢谢参与!您的真实评价是我们改进的动力~
上传资源赚钱or赚积分
最新推荐
脉冲神经网络的监督学习算法研究综述 50积分/C币 立即下载
1/0