Modeling,Dynamics Analysis And Circuit Implementation Of Memristor Neural Network System

Memristors have high integration,nonlinearity and synaptic plasticity,which are considered as the most promising electronic components to simulate biological synaptic information.Studies have shown that changes in the structure of neurons and synapses can affect the memory properties of organisms.Therefore,in the context of neurology and physics,using the synaptic plasticity of memristors to replace the synapses of neurons to establish memristor neural network models has been studied by many scholars.Compared with the general neural network system,the memristor neural network is closer to the actual nervous system in terms of dynamic characteristics.There are rich firing patterns in the biological nervous system.The memristor neural network simulated by memristor can reflect specific functions,such as periodicity or chaos and cluster discharge,which helps people better understand the dynamic characteristics of neurons.This paper focuses on amnestic neural networks.Based on the relevant literature on amnestic neural networks,a new amnestic and neural network model is constructed and its dynamical properties are analysed,and the specific innovations and research work are summarised as follows.(1)A new memristor neural network system with synaptic crosstalk effect was constructed using a hyperbolic tangent amnestic model combined with Hopfield neural network.The system crosstalk effect is classified into three cases,and the changes in the dynamical behaviour of the system under the crosstalk effect are revealed by analysing the system boundedness,equilibrium point,bifurcation diagram,Lyapunov exponential spectrum,attraction domain distribution and phase diagram.By calculating the average Hamiltonian energy change,the phenomenon of system coexistence and multi-stability is revealed from the energy perspective.Finally,the simulation circuit of the system is constructed based on PSIM and the correctness of the numerical simulation of the system is experimentally verified.(2)The proposed hyperbolic tangent memristor model is modified,and a new local active memristor model is constructed to analyze its instability,bistability and local active properties.Then,a new local active memristor synaptic coupling HR neuron model was constructed by using the proposed memristor.By analyzing the coupling effects on the dynamics of the system,it was found that the system has a variety of states,such as periodic discharge,intermittent chaotic discharge and resting state,among which the intermittent chaotic discharge state better reflects the actual activity changes of the biological neurons.In addition,the attraction domain distribution and phase diagram are drawn,and it is found that the system appears giant stability phenomenon by changing the coupling strength.Through Hamiltonian energy analysis,the dynamic changes of the system are explained.Finally,the correctness of the numerical simulation is verified by PSIM circuit simulation.