• 北大核心期刊(《中文核心期刊要目总览》2017版)
  • 中国科技核心期刊(中国科技论文统计源期刊)
  • JST 日本科学技术振兴机构数据库(日)收录期刊

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

具有混合奖惩信号的脉冲时间依赖可塑性算法

陈运享 冯忍 陈云华

陈运享, 冯忍, 陈云华. 具有混合奖惩信号的脉冲时间依赖可塑性算法[J]. 微电子学与计算机, 2022, 39(9): 20-25. doi: 10.19304/J.ISSN1000-7180.2022.0108
引用本文: 陈运享, 冯忍, 陈云华. 具有混合奖惩信号的脉冲时间依赖可塑性算法[J]. 微电子学与计算机, 2022, 39(9): 20-25. doi: 10.19304/J.ISSN1000-7180.2022.0108
CHEN Yunxiang, FENG Ren, CHEN Yunhua. Spiking timing dependent plasticity algorithm with mixed reward-modulated signals[J]. Microelectronics & Computer, 2022, 39(9): 20-25. doi: 10.19304/J.ISSN1000-7180.2022.0108
Citation: CHEN Yunxiang, FENG Ren, CHEN Yunhua. Spiking timing dependent plasticity algorithm with mixed reward-modulated signals[J]. Microelectronics & Computer, 2022, 39(9): 20-25. doi: 10.19304/J.ISSN1000-7180.2022.0108

具有混合奖惩信号的脉冲时间依赖可塑性算法

doi: 10.19304/J.ISSN1000-7180.2022.0108
基金项目: 

广东省自然科学基金 2021A1515012233

详细信息
    作者简介:

    陈运享  男,(1994-),硕士研究生.研究方向为类脑计算、深度学习

    冯忍  男,(1995-),硕士研究生.研究方向为类脑计算、深度学

    通讯作者:

    陈云华(通讯作者)  女,(1977-),博士,副教授.研究方向为类脑计算、深度学习、机器学习. E-mail: yhchen@gdut.edu.cn

  • 中图分类号: TP391

Spiking timing dependent plasticity algorithm with mixed reward-modulated signals

  • 摘要:

    近年来,具有生理学基础的脉冲时间依赖可塑性(Spiking Timing-Dependent Plasticity,STDP)规则在脉冲神经网络中得到了越来越多的应用.由STDP规则和奖惩机制相结合的R-STDP(reward-modulated STDP)学习算法在改善脉冲神经网络的性能上有良好的效果.但R-STDP算法在训练多层脉冲神经网络时,仍存在反馈信号仅作用于网络末层、中间层无法获得有用奖惩信号.为此,利用自编码器的无监督特性,提出一种具有混合奖惩信号的MR-STDP(Mix Reward-modulated STDP)算法.在中间层中增加重构层以够建基于卷积自编码器的奖惩信号因子模型,通过比较卷积层和重构层的神经元脉冲发放时间,获取中间层网络权重调整的指导因子信号.指导因子信号是对比层间自编码器的输入层与重构层的相同位置神经元所发放的脉冲序列相似性度量指标,并将其与R-STDP相结合,使得中间层能够获得权重指导信号.在MNIST和COVID-19 CT数据集上的实验结果表明,该方法取得了比R-STDP更高的精度,且中间层网络的学习效率大幅提高.

     

  • 图 1  网络结构图

    Figure 1.  Network structure diagram

    图 2  卷积自编码器

    Figure 2.  convolution autoencoder

    图 3  MNIST和COVID-19数据集例图

    Figure 3.  Sample plots of MNIST and COVID datasets

    图 4  ε值与准确率统计图

    Figure 4.  ε value and accuracy statistics

    图 5  MNIST数据集精度变化曲线图

    Figure 5.  The accuracy change of MNIST dataset

    图 6  ε值与准确率统计图

    Figure 6.  The ε value and accuracy statistics

    图 7  COVID-19数据集上精度和损失变化曲线

    Figure 7.  Change curves of accuracy and loss on the COVID-19 dataset

    表  1  网络参数设置

    Table  1.   Parameter setting in the network

    参数 C1 C2 C3
    AP+ 0.004 0.004 0.004
    AP -0.003 -0.003 -0.003
    Ar+ -0.004 -0.004 -0.0005
    Ar 0.003 0.003 0.004
    τ+ 4 4 -
    τ 2 2 -
    阈值 15 10
    γ 1 1 0
    重构层阈值 30 20 -
    下载: 导出CSV

    表  2  MNIST数据集上的精度比较

    Table  2.   Accuracy comparison on MNIST dataset

    方法 浅层奖惩信号 网络结构 学习方法 准确率/%
    Peter[4] 不适用 单层卷积 A-STDP 95.0
    Amir[6] 不适用 多层全连接 BP-STDP 97.2
    Milad[16] × 多层卷积 R-STDP 97.2
    Chanky[19] × 多层卷积 STDP 91.1
    本文 多层卷积 MR-STDP 97.8
    下载: 导出CSV

    表  3  网络模型形同中间层层网络训练耗时比较

    Table  3.   Comparison of middle layer training time

    方法 C1层/EPOCH C2层/EPOCH 耗时/S 准确率/%
    MILAD[16] 2 4 1472 97.2
    本文 1 1 376 97.8
    下载: 导出CSV

    表  4  COVID-19数据集上的精度比较

    Table  4.   Accuracy comparison on COVID-19 data sets

    方法 学习算法 网络结构 浅层奖惩信号 准确率/%
    MILAD[17] R-STDP 多层卷积结构 × 60.82
    AVISHEK[20] STDP+R-STDP 多层卷积结构 × 71.00
    本文 MR-STDP 多层卷积结构 81.89
    下载: 导出CSV

    表  5  中间层网络训练耗时比较

    Table  5.   Comparison of middle layer training time

    方法 C1层/EPOCH C2层/EPOCH 耗时/S 准确率/%
    AVISHEK[20] 15 15 2 528 70.00
    本文 5 10 904 81.89
    下载: 导出CSV
  • [1] TAVANAEI A, GHODRATI M, KHERADPISHEH S R, et al. Deep learning in spiking neural networks[J]. Neural Networks, 2019(111): 47-63. DOI: 10.1016/j.neunet.2018.12.002.
    [2] PFEIFFER M, PFEIL T. Deep learning with spiking neurons: opportunities and challenges[J]. Frontiers in Neuroscience, 2018(12): 774. DOI: 10.3389/fnins.2018.00774.
    [3] MASQUELIER T, THORPE S J. Unsupervised learning of visual features through spike timing dependent plasticity[J]. PLoS Computational Biology, 2007, 3(2): e31. DOI: 10.1371/journal.pcbi.0030031.
    [4] DIEHL P U, COOK M. Unsupervised learning of digit recognition using spike-timing-dependent plasticity[J]. Frontiers in Computational Neuroscience, 2015(9): 99. DOI: 10.3389/fncom.2015.00099.
    [5] KHERADPISHEH S R, GANJTABESH M, THORPE S J, et al. STDP-based spiking deep convolutional neural networks for object recognition[J]. Neural Networks, 2018(99): 56-67. DOI: 10.1016/j.neunet.2017.12.005.
    [6] TAVANAEI A, MAIDA A. BP-STDP: approximating backpropagation using spike timing dependent plasticity[J]. Neurocomputing, 2019(330): 39-47. DOI: 10.1016/j.neucom.2018.11.014.
    [7] LEE C, SRINIVASAN G, PANDA P, et al. Deep spiking convolutional neural network trained with unsupervised spike-timing-dependent plasticity[J]. IEEE Transactions on Cognitive and Developmental Systems, 2019, 11(3): 384-394. DOI: 10.1109/TCDS.2018.2833071.
    [8] KABLE J W, GLIMCHER P W. The neurobiology of decision: consensus and controversy[J]. Neuron, 2009, 63(6): 733-745. DOI: 10.1016/j.neuron.2009.09.003.
    [9] SERRE T, WOLF L, BILESCHI S, et al. Robustobject recognition with cortex-like mechanisms[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2007, 29(3): 411-426. DOI: 10.1109/TPAMI.2007.56.
    [10] LEE D, SEO H, JUNG M W. Neural basis of reinforcement learning and decision making[J]. Annual Review of Neuroscience, 2012(35): 287-308. DOI: 10.1146/annurev-neuro-062111-150512.
    [11] FRMAUX N, GERSTNER W. Neuromodulated spike-timing-dependent plasticity, and theory of three-factor learning rules[J]. Frontiers in Neural Circuits, 2016(9): 85. DOI: 10.3389/fncir.2015.00085.
    [12] IZHIKEVICH E M. Solving the distal reward problem through linkage of STDP and dopamine signaling[J]. Cerebral Cortex, 2007, 17(10): 2443-2452. DOI: 10.1093/cercor/bhl152.
    [13] FARRIES M A, FAIRHALL A L. Reinforcement learning with modulated spike timing-dependent synaptic plasticity[J]. Journal of Neurophysiology, 2007, 98(6): 3648-3665. DOI: 10.1152/jn.00364.2007.
    [14] FLORIAN R V. Reinforcement learning through modulation of spike-timing-dependent synaptic plasticity[J] Neural Computation, 2007, 19(6): 1468-1502. DOI: 10.1162/neco.2007.19.6.1468.
    [15] MOZAFARI M, KHERADPISHEH S R, MASQUELIER T, et al. First-spike-based visual categorization using reward-modulated STDP[J]. IEEE Transactions on Neural Networks and Learning Systems, 2018, 29(12): 6178-6190. DOI: 10.1109/TNNLS.2018.2826721.
    [16] MOZAFARI M, GANJTABESH M, NOWZARI-DALINI A, et al. Bio-inspired digit recognition using reward-modulated spike-timing-dependent plasticity in deep convolutional networks[J]. Pattern Recognition, 2019(94): 87-95. DOI: 10.1016/j.patcog.2019.05.015.
    [17] THORPE S, FIZE D, MARLOT C. Speed of processing in the human visual system[J]. Nature, 1996, 381(6582): 520-522. DOI: 10.1038/381520a0.
    [18] MOZAFARI M, GANJTABESH M, NOWZARI-DALINI A, et al. SpykeTorch: efficient simulation of convolutional spiking neural networks with at most one spike per neuron[J]. Frontiers in Neuroscience, 2019(13): 625. DOI: 10.3389/fnins.2019.00625.
    [19] LEE C, SRINIVASAN G, PANDA P, et al. Deep spiking convolutional neural network trained with unsupervised spike-timing-dependent plasticity[J]. IEEE Transactions on Cognitive and Developmental Systems, 2019, 11(3): 384-394. DOI: 10.1109/TCDS.2018.2833071.
    [20] GARAIN A, BASU A, GIAMPAOLO F, et al. Detection of COVID-19 from CT scan images: a spiking neural network-based approach[J]. Neural Computing and Applications, 2021, 33(19): 12591-12604. DOI: 10.1007/s00521-021-05910-1.
  • 加载中
图(7) / 表(5)
计量
  • 文章访问数:  71
  • HTML全文浏览量:  61
  • PDF下载量:  14
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-02-18
  • 修回日期:  2022-04-05
  • 网络出版日期:  2022-09-15

目录

    /

    返回文章
    返回