Design And Application Of Memristor Neural Networks With Time Delay

Memristor,a nonlinear basic circuit element that characterizes the relationship between charge and magnetic flux,whose resistance value can change with input current or voltage.Neural network,a mathematical algorithm model of distributed parallel information processing,achieves the purpose of processing information by adjusting the interconnected relationship between a large number of internal nodes.In analog circuit implementations of neural networks,time delay is usually unavoidable due to the limited switching speed of amplifiers,and it is the main cause of instability and poor performance of networks.Therefore,it will be of great theoretical significance and practical values to study the dynamic behavior of neural network with time delay.In this paper,research work mainly focuses on that construct and analyze a newly memristive time-delay neural network to apply to Pavlovian associative memory.Further research and exploration of the dynamic characteristics of the memristive time-delay neural network and its dynamics in associative memory to help advance the research and development of memristive time-delay neural networks.The main results of this paper are as follows,(1)A memristive time-delay recurrent neural network(RNN)model is constructed based on the hyperbolic tangent memristor.The equilibrium point and instability of the neural network model are discussed.By adjusting the size of the internal parameters of the memristor,the coupling coefficient of the memristor and the time delay,the mixed effects of the internal parameters of the memristor,the coupling coefficient of the memristor and the time-delay on the chaotic dynamics of the memristor time-delay RNN are analyzed in detail.(2)The circuit experimental model of the memristive time-delay RNN is built based on the neuron circuit,the hyperbolic tangent function circuit and the memristor circuit.The complex dynamics of the memristive time-delay RNN are simulated by experiments with different internal parameters of the memristor,coupling coefficients and time delays,which verifies the results of theoretical analysis and numerical simulation.(3)An equivalent circuit model of the hyperbolic tangent memristor with threshold is proposed based on the hyperbolic tangent memristor,and verified by simulation.Inspired by the basis of neural networks and biological phenomena,a multifunctional Pavlov associative circuit model is constructed based on the hyperbolic tangent memristor with threshold,which innovatively added a time-delay input design.Through three main circuit modules,that is,salivation synapse module,time-delay input module and rate control module,the whole circuit can realize Pavlov associative memory with functions of time-delay learning,consolidation learning and long-term memory.