计算机科学 ›› 2024, Vol. 51 ›› Issue (11A): 240300191-6.doi: 10.11896/jsjkx.240300191

• 智能计算 • 上一篇    下一篇

基于多子网络预训练的脉冲神经网络分类模型

卓明松, 莫凌飞   

  1. 东南大学仪器科学与工程学院 南京 210096
  • 出版日期:2024-11-16 发布日期:2024-11-13
  • 通讯作者: 莫凌飞(lfmo@seu.edu.cn)
  • 作者简介:(220213666@seu.edu.cn)
  • 基金资助:
    江苏省高校青蓝工程

Spiking Neural Network Classification Model Based on Multi-subnetwork Pre-training

ZHUO Mingsong, MO Lingfei   

  1. School of Instrument Science and Engineering,Southeast University,Nanjing 210096,China
  • Online:2024-11-16 Published:2024-11-13
  • About author:ZHUO Mingsong,born in 1999,postgraduate.His main research interest is spiking neural network learning algorithms and their applications.
    MO Lingfei,born in 1981,Ph.D,asso-ciate professor,Ph.D supervisor.His main research interests includeneuromorphic intelligence and neuromorphic perception,and intelligent perception of the Internet of Things.
  • Supported by:
    Blue Project of Jiangsu Province.

摘要: 脉冲神经网络(Spiking Neural Network,SNN)被认为是最符合生物大脑机制的类脑计算模型,凭借其事件驱动、高能效、可解释等特点吸引了越来越多的研究关注。然而,由于脉冲的二值输出与不可微分性,SNN的训练方法仍存在一定空缺。于是借鉴皮层记忆单元通过局部网络存储记忆信息的方式,提出一种基于多子网络预训练的脉冲神经网络分类方法。该方法使用样本标签信息优化了脉冲序列特征提取过程,采用改进的脉冲时间依赖可塑性学习规则预训练多个单类别特征提取子网络,并将预训练后的子网络进行无监督特征融合,有效提高了网络的特征分类能力。此外,在权重可视化与t-SNE可视化工具的帮助下,分析了方法的有效性。所提方法在MNIST与Fashion-MNIST数据集上分别取得了97.40%与88.81%的分类准确度。

关键词: 脉冲神经网络, 脉冲时间依赖可塑性, 单类别特征提取子网络, 无监督特征融合, 类脑计算

Abstract: Spiking neural network(SNN) is widely regarded as the most biologically plausible model,aligning closely with the mechanisms of the biological brain.It has garnered increasing research attention due to its event-driven nature,high energy efficiency,and interpretability.However,the training methods of SNN still have some limitations due to the binary output and non-differentiability of the spike.This paper proposes a SNN classification method based on multi-subnetwork pre-training,which draws inspiration from the way cortical memory units store memory information through local networks.This approach leverages sample label information to optimize the feature extraction process,employs enhanced spike-timing-dependent-plasticity learning rules for pre-training multiple single-class feature extraction subnetworks,and conducts unsupervised feature fusion on these pre-trained subnetworks to effectively enhance the network's feature classification capability.Furthermore,the effectiveness of this method is analyzed through weight visualization and t-SNE visualization tools.Finally,the classification accuracy of 97.40% and 88.81% is achieved on MNIST and Fashion MNIST datasets,respectively.

Key words: Spiking neural network, Spike-timing-dependent-plasticity, Single-class feature extraction subnetworks, Unsupervised feature fusion, Brain-like computing

中图分类号: 

  • TP389.1
[1]KRIZHEVSKY A,SUTSKEVER I,HINTONG E.Imagenetclassification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[2]KALCHBRENNER N,GREFENSTETTE E,BLUNSOM P.A Convolutional Neural Network for Modelling Sentences[C]//Annual meeting of the Association for Computational Linguistics.2014.
[3]SILVER D,SCHRITTWIESER J,SIMONYAN K,et al.Mastering the game of go without human knowledge[J].Nature,2017,550(7676):354-359.
[4]WANG X Q,ZENG Y H,HAN D M,et al.Brain like computing based on spiking neural networks [J].Journal of Beijing Institute of Technology,2019,45(12):1277-1286.
[5]MAASS W.Networks of spiking neurons:the third generation of neural network models[J].Neural Networks,1997,10(9):1659-1671.
[6]CAPORALE N,DAN Y.Spike timing-dependent plasticity:aHebbian learning rule[J].Annu.Rev.Neurosci.,2008,31:25-46.
[7]LESTER D,FURBER S.Spinnaker:Distributed computer engineering for neuromorphics[M]//Neural Nets WIRN11.IOS Press,2011:324-331.
[8]PEI J,DENG L,SONG S,et al.Towards artificial general intelligence with hybrid Tianjic chip architecture[J].Nature,2019,572(7767):106-111.
[9]DAVIES M,SRINIVASA N,LIN T H,et al.Loihi:A neuromorphic manycore processor with on-chip learning[J].IEEE Micro,2018,38(1):82-99.
[10]MA D,SHEN J,GU Z,et al.Darwin:A neuromorphic hardware co-processor based on spiking neural networks[J].Journal of Systems Architecture,2017,77:43-51.
[11]BOUVIER M,VALENTIAN A,MESQUIDA T,et al.Spiking neural networks hardware implementations and challenges:A survey[J].ACM Journal on Emerging Technologies in Computing Systems(JETC),2019,15(2):1-35.
[12]DIEHL P U,COOK M.Unsupervised learning of digit recognition using spike-timing-dependent plasticity[J].Frontiers in Computational Neuroscience,2015,9:99.
[13]KOHONEN T.The self-organizing map[J].Proceedings of the IEEE,1990,78(9):1464-1480.
[14]TEYLER T J,RUDY J W.The hippocampal indexing theoryand episodic memory:updating the index[J].Hippocampus,2007,17(12):1158-1169.
[15]PONULAK F.ReSuMe-new supervised learning method for spiking neural networks.Institute of Control and Information Engineering,Poznoń University of Technology[R].Tech.rep,2005.
[16]ZHANG T,ZENG Y,ZHAO D,et al.A plasticity-centric ap-proach to train the non-differential spiking neural networks[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2018.
[17]KOCH C,SEGEV I.Methods in neuronal modeling:from ions to networks[M].MIT press,1998.
[18]HODGKIN A L,HUXLEYA F.A quantitative description ofmembrane current and its application to conduction and excitation in nerve[J].The Journal of Physiology,1952,117(4):500.
[19]KRUG M,MÜLLER-WELDE P,WAGNERM,et al.Functional plasticity in two afferent systems of the granule cells in the rat dentate area:frequency-related changes,long-term potentiation and heterosynaptic depression[J].Brain Research,1985,360(1/2):264-272.
[20]QUERLIOZ D,BICHLER O,DOLLFUS P,et al.Immunity to device variations in a spiking neural network with memristive nanodevices[J].IEEE Transactions on Nanotechnology,2013,12(3):288-295.
[21]HAO Y,HUANG X,DONG M,et al.A biologically plausiblesupervised learning method for spiking neural networks using the symmetric STDP rule[J].Neural Networks,2020,121:387-395.
[22]MO L,WANG G,LONG E,et al.ALSA:Associative Learning Based Supervised Learning Algorithm for SNN[J].Frontiers in Neuroscience,2022,16:838832.
[23]LIU F,ZHAO W,CHEN Y,et al.SSTDP:Supervised spike ti-ming dependent plasticity for efficient spiking neural network training[J].Frontiers in Neuroscience,2021,15:756876.
[24]TAVANAEI A,MAIDA A.BP-STDP:Approximating back-propagation using spike timing dependent plasticity[J].Neurocomputing,2019,330:39-47.
[25]PUTRA R V W,SHAFIQUE M.Fspinn:An optimizationframework for memory-efficient and energy-efficient spiking neural networks[J].IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems,2020,39(11):3601-3613.
[26]RASTOGI M,LU S,ISLAM N,et al.On the self-repair role of astrocytes in STDP enabled unsupervised SNNs[J].Frontiers in Neuroscience,2021,14:603796.
[27]SHI M,ZHANG T,ZENG Y.A curiosity-based learning method for spiking neural networks[J].Frontiers in Computational Neuroscience,2020,14:7.
[28]ZHAO D,ZENG Y,ZHANG T,et al.GLSNN:A multi-layer spiking neural network based on global feedback alignment and local STDP plasticity[J].Frontiers in Computational Neuroscience,2020,14:576841.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!