一种新型绝对值忆阻耦合自突触Hopfield神经网络的动力学分析及其电路实现

Dynamics analysis and circuit implementation of a novel absolute value memristor coupled self-synaptic Hopfield neural network

神经元作为大脑基本的组成单元能够产生复杂的动力学行为。目前大部分的研究是关于两个神经元系统的忆阻耦合突触,而忆阻耦合自突触权重的单神经元模型的研究相对较少。本文提出了绝对值忆阻耦合自突触权值的Hopfield神经网络(HNN)模型,以自耦合权重作为唯一的调节参数。利用基本的动力学分析方法,讨论了不同耦合强度下系统的动力学行为,研究了不同初始值下对称吸引子的共存行为。结果表明,这些丰富的非线性动力学行为包括周期倍增分岔、混沌、周期窗和对称自激吸引子共存。最后,通过PSpice仿真验证了所提出的忆阻HNN的理论分析结果的正确性。

As the fundamental building blocks of brain,neurons can produce complex dynamic behaviors.For memristor based neural network,most research is about the memristor coupling of synapses between two neuron systems.A very few have been done with the single neuron model of memristor coupling from the weight of synapses.A Hopfield neural network(HNN)model was proposed with an absolute memristor coupled self-synaptic weights,where the self-coupling weight was used as the only regulating parameter.By using the fundamental dynamic analysis method,the dynamic behavior was discussed under different coupling strength,and the attractor coexistence behavior was studied under different initial values.The results show that these abundant nonlinear dynamical behaviors include periodic doubling bifurcation,chaos,periodic window and symmetric self-excited attractor coexistence.Finally,theoretical analysis was verified by PSpice simulation for the proposed memristor HNN.The experimental results validate the theoretical analysis as well.