计算物理 ›› 2022, Vol. 39 ›› Issue (2): 244-252.DOI: 10.19596/j.cnki.1001-246x.8386

• 研究论文 • 上一篇    

忆阻Hopfield神经网络动力学分析及其电路实现

唐利红1(), 贺宗梅1, 姚延立2   

  1. 1. 长沙民政职业技术学院软件学院, 湖南 长沙 410007
    2. 滨州学院航空工程学院, 山东 滨州 256600
  • 收稿日期:2021-04-27 出版日期:2022-03-25 发布日期:2022-06-24
  • 作者简介:

    唐利红,E-mail:

  • 基金资助:
    湖南省教育厅一般项目(20c0101)

Dynamical Analysis and Circuit Implementation of a Memristive Hopfield Neural Network

Lihong TANG1(), Zongmei HE1, Yanli YAO2   

  1. 1. School of Software, Changsha Social Work College, Changsha, Hunan 410007, China
    2. College of Aeronautical Engineering, Binzhou University, Binzhou, Shandong 256600, China
  • Received:2021-04-27 Online:2022-03-25 Published:2022-06-24

摘要:

提出一种超多稳态忆阻Hopfield神经网络, 它仅包含3个神经元和一个多稳态忆阻突触。从理论上分析神经网络的耗散性和平衡点的稳定性, 并利用分岔图、李雅普诺夫指数谱和相位图等数值方法分析不同忆阻突触耦合强度对神经网络动力学的影响。网络参数固定时, 揭示与初始状态值密切相关的超多稳态性动力学行为。最后, 设计忆阻Hopfield神经网络的模拟等效电路, 并通过PSIM电路仿真验证MATLAB数值仿真结果。

关键词: 忆阻器, 神经网络, 多稳态, 混沌, 非线性电路

Abstract:

An extremely multistable memristive Hopfield neural network (HNN) is proposed which includes only three neurons and one multistable memristor synapse. Dissipativity and stability of equilibrium points are theoretically analyzed, and influence of memristive synapse-coupled strengths on dynamics in the memristive neural network is analyzed with numerical methods such as bifurcation diagrams, Lyapunov exponents, and phase plots. As network parameters are fixed, dynamical behavior of extreme multistability related to initial states is revealed. Finally, analog equivalent circuit of the memristive HNN is designed, and MATLAB numerical simulation results are verified with PSIM circuit simulation.

Key words: memristor, neural network, multistable, chaos, nonlinear circuit