一种绝对值忆阻Hopfield神经网络的动力学分析与其实现  

Dynamic analysis and implementation of an absolute memristor Hopfield neural network

作　　者：李旭鑫 邱达 陈世强 罗敏 刘嵩 LI Xuxin;QIU Da;CHEN Shiqiang;LUO Min;LIU Song(College of Intelligent Systems Science and Engineering,Hubei Minzu University,Enshi445000,Hubei Province,China)

机构地区：[1]湖北民族大学智能科学与工程学院,湖北恩施445000

出　　处：《电子元件与材料》2024年第10期1264-1273,共10页Electronic Components And Materials

基　　金：湖北省自然科学基金(2024AFD068)。

摘　　要：Hopfield神经网络(HNN)是一种类脑神经网络,可以表现出丰富的动态行为,尤其是混沌。为了探究这些丰富的动力学现象,研究了一种具有忆阻突触权重的Hopfield神经网络模型,该HNN模型具有三个平衡点。通过动力学地图、分岔图、Lyapunov指数谱、相轨图和时序图分析了系统的动力学行为,研究了忆阻突触耦合强度和初始条件对系统动力学行为的影响。实验结果表明,构建的Hopfield神经网络不仅会随着忆阻强度的变化出现倍周期分岔、混沌、周期窗口和对称共存吸引子,还能随着状态初值的变化产生多种类型的非对称共存吸引子(例如周期和混沌共存吸引子以及单涡卷和双涡卷共存吸引子),同时还发现了暂态混沌现象。最后,通过FPGA技术实现了系统的数字电路,验证了所提出系统的正确性和可实现性。The Hopfield neural network is a brain-like neural network that can exhibit rich dynamical behaviors,especially chaos.To explore these rich dynamical phenomena,a Hopfield neural network model was investigated with memristive synaptic weights,which has three equilibrium points.The dynamic behavior of the system was analyzed by dynamic maps,bifurcation diagrams,Lyapunov exponent spectrum,phase trajectory diagrams,and time series diagrams.Investigation was also performed to study the effect of memristive synaptic coupling strength and initial conditions on the system's dynamic behavior.The experimental results show that the constructed Hopfield neural network can not only exhibits period-doubling bifurcation,chaotic and periodic windows,and symmetric co-existing attractors under different memristive strength,but also produces various types of asymmetric coexistence attractors with the change of the initial state value(such as periodic and chaotic coexistence attractors,single-scroll and double-scroll coexistence attractors),and also discovers transient chaos phenomena.Finally,the digital circuit of the system was implemented by FPGA technology,verifying the correctness and realizability of the proposed system.

关 键 词：HOPFIELD神经网络 忆阻突触 共存吸引子 瞬态混沌 FPGA实现 

分 类 号：TN711.4[电子电信—电路与系统] TM132[电气工程—电工理论与新技术]