忆阻自突触Hopfield神经网络的动力学分析与电路仿真  

Memristor-based self-synaptic Hopfield neural network:dynamic analysis and circuit simulation

作　　者：张学丰 彭良玉 彭代鑫 ZHANG Xuefeng;PENG Liangyu;PENG Daixin(School of Physics and Electronics,Hunan Normal University,Changsha 410081,China)

机构地区：[1]湖南师范大学物理与电子科学学院,湖南长沙410081

出　　处：《电子元件与材料》2022年第3期315-322,共8页Electronic Components And Materials

摘　　要：通过理论分析和仿真的方法,研究了一种4维忆阻自突触Hopfiled神经网络(以下简称MAHNN)的动力学行为及其仿真电路的实现。首先,分析了MAHNN能够产生复杂动力学行为的基本条件。其次,利用常规的动力学分析方法如分岔图、李氏指数、相轨迹图和时域图,分析了它的动力学行为。MATLAB数值仿真结果揭示了MAHNN存在与忆阻控制参数有关的特殊动力学行为。最后,采用改进型模块化电路设计方法,设计了MAHNN的模拟等效电路。Multisim电路仿真结果与MATLAB数值仿真结果一致,验证了忆阻自突触Hopfield神经网络理论设计的正确性。By theoretical analysis and simulation,the dynamic behavior of a 4-dimensional memristor-based self-synaptic Hopfield Neural Network(MSHNN)and its simulation implementation were studied.Firstly,the basic conditions for the MSHNN to produce complex dynamic behavior were analyzed.Secondly,the dynamic behavior of the MSHNN was analyzed by using conventional dynamic analysis methods,such as bifurcation diagram,Lyapunov exponents,phase portraits and time domain diagram.MATLAB numerical simulation results reveal that the MSHNN has special dynamic behavior related to memristor control parameter.Finally,the analog equivalent circuit of the MAHNN was designed by using the improved modular circuit design method.Multisim simulation results are consistent with the numerical simulation results of MATLAB,which verifies the correctness of the theoretical design of the MSHNN.

关 键 词：忆阻器 自突触 HOPFIELD神经网络 动力学行为 混沌吸引子 电路设计 

分 类 号：TN722.7[电子电信—电路与系统] TM132[电气工程—电工理论与新技术]