• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

计算机工程与科学 ›› 2020, Vol. 42 ›› Issue (07): 1325-1330.doi: 10.3969/j.issn.1007-130X.2020.07.023

• 数据挖掘与人工智能 • 上一篇    

不同重连概率的小世界脉冲神经网络抗扰功能研究

郭磊1,2,冯海1,2,石洪溢1,2   

  1. (1.天津市生物电工与智能健康重点实验室(河北工业大学),天津 300130;

    2.省部共建电工装备可靠性与智能化国家重点实验室(河北工业大学),天津 300130)


  • 收稿日期:2019-09-25 修回日期:2020-02-27 接受日期:2020-07-25 出版日期:2020-07-25 发布日期:2020-07-27
  • 基金资助:
    国家自然科学基金(61571180)

Anti-interference ability of small world spiking neural networks with different rewiring probabilities

GUO Lei1,2,FENG Hai1,2,SHI Hong-yi1,2   

  1. (1.Tianjin Key Laboratory of Bioelectromagnetic Technology and Intelligent Health,
    Hebei University of Technology,Tianjin 300130;

    2.State Key Laboratory of Reliability and Intelligence of Electrical Equipment,
    Hebei University of Technology,Tianjin 300130,China)

  • Received:2019-09-25 Revised:2020-02-27 Accepted:2020-07-25 Online:2020-07-25 Published:2020-07-27

摘要: 现今各种电磁干扰对电子系统造成的不良影响越来越严重,传统防护方式的局限性日益凸显。电磁仿生学由此被提出,目的是通过借鉴生物体的自适应抗扰的优良特性,以期建立新的防护模式。构建了以Izhikevich神经元模型为节点,兴奋性和抑制性突触可塑性模型共同调节基于小世界网络拓扑的小世界脉冲神经网络;基于复杂网络理论对比分析了不同重连概率的小世界网络的拓扑特性;对比分析了不同重连概率的小世界脉冲神经网络在高斯白噪声刺激下的抗扰功能。实验结果表明:小世界网络的平均路径长度和全局效率值受重连概率的影响较小,平均聚类系数和小世界属性受重连概率的影响较大;构建的不同重连概率的脉冲神经网络均具有一定抗扰功能且高聚类系数和低平均路径长度显著的小世界脉冲神经网络抗扰功能最优。


关键词: 小世界网络, 脉冲神经网络, 重连概率, 突触可塑性, 抗扰功能

Abstract: Nowadays, the adverse effects of various electromagnetic interferences on electronic systems are becoming more serious today, and the shortcomings of traditional anti-electromagnetic interfe- rence methods are increasingly prominent. Electromagnetic bionics proposes to establish a new protection mode based on the bionic model by referring to the excellent characteristics of adaptive anti-interference of organisms. The small-world spiking neural network with the Izhikevich model and the plasticity of excitatory and inhibitory is constructed. Based on complex network theory, topology characteristics of small world networks with different rewiring probability are compared. The anti-interference ability of small world spiking neural networks with different rewiring probability under Gaussian white noise are compared. The experimental results show that average path length and global efficiency of the small world network are less affected by the rewiring probability, and average clustering coefficient and small world property are greatly affected by the rewiring probability. The small world spiking neural network has certain anti-interference ability, and the network with high clustering coefficient and low average path length has the best anti-interference ability.



Key words: small world network, spiking neural network, rewiring probability, synaptic plasticity, anti-interference