电子学报 ›› 2016, Vol. 44 ›› Issue (12): 2877-2886.DOI: 10.3969/j.issn.0372-2112.2016.12.010

• 学术论文 • 上一篇    下一篇

基于脉冲序列核的脉冲神经元监督学习算法

蔺想红, 王向文, 党小超   

  1. 西北师范大学计算机科学与工程学院, 甘肃兰州 730070
  • 收稿日期:2015-04-07 修回日期:2015-07-01 出版日期:2016-12-25
    • 通讯作者:
    • 党小超
    • 作者简介:
    • 蔺想红,男,1976年1月生于甘肃天水.2009年获哈尔滨工业大学计算机应用技术专业博士学位,现任西北师范大学计算机科学与工程学院教授,硕士生导师.研究方向为神经网络、进化计算、人工生命、图像处理.E-mail:linxh@nwnu.edu.cn;王向文,男,1991年3月生于甘肃天水.2015年获西北师范大学软件工程专业硕士学位.研究方向为神经网络、机器学习.E-mail:wangxiangwen2@163.com
    • 基金资助:
    • 国家自然科学基金 (No.61165002,No.61363059); 甘肃省自然科学基金 (No.1506RJZA127); 甘肃省高等学校科研项目 (No.2015A-013)

A New Supervised Learning Algorithm for Spiking Neurons Based on Spike Train Kernels

LIN Xiang-hong, WANG Xiang-wen, DANG Xiao-chao   

  1. School of Computer Science and Engineering, Northwest Normal University, Lanzhou, Gansu 730070, China
  • Received:2015-04-07 Revised:2015-07-01 Online:2016-12-25 Published:2016-12-25

摘要:

脉冲神经元应用脉冲时间编码神经信息,监督学习的目标是对于给定的突触输入产生任意的期望脉冲序列.但由于神经元脉冲发放过程的不连续性,构建高效的脉冲神经元监督学习算法非常困难,同时也是该研究领域的重要问题.基于脉冲序列的核函数定义,提出了一种新的脉冲神经元监督学习算法,特点是应用脉冲序列核构造多脉冲误差函数和对应的突触学习规则,并通过神经元的实际脉冲发放频率自适应地调整学习率.将该算法用于脉冲序列的学习任务,期望脉冲序列采用Poisson过程或线性方法编码,并分析了不同的核函数对算法学习性能的影响.实验结果表明该算法具有较高的学习精度和良好的适应能力,在处理复杂的时空脉冲模式学习问题时十分有效.

关键词: 脉冲神经元, 监督学习, 脉冲序列核, 内积, 脉冲序列学习

Abstract:

The purpose of supervised learning with temporal encoding for spiking neurons is to make the neurons emit arbitrary spike trains in response to given synaptic inputs.However,due to the discontinuity in the spike process,the formulation of efficient supervised learning algorithms for spiking neurons is difficult and remains an important problem in the research area.Based on the definition of kernel functions for spike trains,this paper proposes a new supervised learning algorithm for spiking neurons with temporal encoding.The learning rule for synapses is developed by constructing the multiple spikes error function using spike train kernels,and its learning rate is adaptively adjusted according to the actual firing rate of spiking neurons during learning.The proposed algorithm is successfully applied to various spike trains learning tasks,in which the desired spike trains are encoded by Poisson process or linear method.Furthermore,the effect of different kernels on the performance of the learning algorithm is also analyzed.The experiment results show that our proposed method has higher learning accuracy and flexibility than the existing learning methods,so it is effective for solving complex spatio-temporal spike pattern learning problems.

Key words: spiking neuron, supervised learning, spike train kernel, inner product, spike train learning

中图分类号: