Research Of Memrisive Hopfield Neural Network

The brain works in chaos,which can be well simulated by Hopfield neural network.It is a common model of the artificial neural computing.The hardware realization of the neural network has always been the key research object of researchers.The synapse weight in the traditional neural network hardware circuit is usually simulated by the unalterable resistors or the CMOS devices.Once the synapse weight is confirmed,it cannot be changed,which hinders the research progress of the neural network hardware circuit.Memristors,a new nonlinear circuit device,become one of the best candidates for simulating synapses due to its adjustable resistance and memory characteristics.In addition,studies have shown that there is the synaptic crosstalk between biological neurons,which will affect the activities and functions of the brain.However,the synaptic crosstalk is seldom considered by the domestic and foreign researchers in the study of neural networks.Therefore,this thesis proposes a newly memristive neural network model,which can simulate the synapse crosstalk.By combining resistors and memristors as synapses,a fourth-order memristive Hopfield neural network is constructed.The dynamic characteristics of these two models are analyzed in detail.Finally,an improved combinatorial algorithm is obtained by combining the transient chaos,introduced into Hopfield neural network,with the variable neighborhood search algorithm,which can solve the classical traveling salesman problem in the combinatorial optimization.The main innovative works of this thesis are as follows,(1)A Hopfield neural network model is constructed based on the definition and characteristics of the memristors,for simulating the synapse crosstalk.The equilibrium points and stabilities of the system are analyzed through the MATLAB software.Moreover,the influences of memristor parameters,crosstalk intensity parameters and the initial value sensitivity on the dynamics behavior of the system are studied in detail.(2)A new fourth-order memristive Hopfield neural network is constructed,by combining the resistors and the memristors as synapse weight of the biological neurons.The complex dynamics of the system are analyzed,including the equilibrium points,stability,phase diagram,time domain diagram,bifurcation diagram and dynamic map,etc.Through theoretical analysis,MATLAB numerical analysis and PSpice circuit simulation experiment,it is proved that the memristive Hopfield neural network system has abundant complex dynamic behaviors,including a variety of types of co-existence attractors,and the highly-sensitive and random chaotic sequence,which can be applied to the optimization calculation and the information encryption.(3)Due to the characteristic that the transient chaos can help the network jump out of the local minimum,thereby stabilizing at the energy minimum,the transient chaos neural network is applied to solve the traveling salesman problem.An improved combined algorithm is proposed by combining the transient chaos with the variable neighborhood search algorithm.The simulation experiment and the data analysis show the advantages of the proposed improved algorithm.