Modeling Analysis And Synchronization Of Memristor Neural Networks

In recent years,memristor devices have achieved rapid development in the fields of physics and materials.As a non-volatile memory,memristors have always been regarded as the best choice for realizing synapses in artificial neural networks.Because of this,the study of memristive neural network has received extensive attention from scholars.In the current research background,the event-triggered control problem of a class of static memristive neural networks(or state-dependent switching neural networks)is studied in this paper,and on this basis,a class of dynamic memristive neural network models is also studied.Then the existing results are extended to the field of stochastic time delay,and the related analysis and control work of the constructed system is carried out.The main contributions are given as follows:1.The dissipative synchronization problem of delayed Markov jump switched neural networks(MJSNNs)under state-dependent switching by the event-triggered gain-scheduling control scheme is studied.By the introduce of a Markov jump model,which is used to depict the random variation wherein the connection of MJSNNs,the issues considered here are more generality.Via constructing suitable Lyapunov-Krasovskii functionals(LKFs)and applying some matrix inequality scaling methods,sufficient conditions for dissipative synchronization of delayed MJSNN are established.According to such criteria,the event-triggered gain-scheduling control scheme is adopted to design a controller with less terminal communication costs.Finally,a numerical example is given to demonstrate the effectiveness of the proposed method.2.A new type of dynamic memristor-based delayed neural networks recently has been paid full attention,where the signal processing is in the charge-flux domain rather than in the current-voltage domain as the standard cellular neural networks or Hopfield neural networks.The most prominent merit of such neural networks is that once the system reaches a steady state,all the basic circuit parameters(currents,voltages and power dissipation)are vanished and the memristors play the role of nonvolatile memories,reserving the processing consequence.Since the existing literature hardly considers the possibility of system parameters changing due to various factors,we extend the semi-Markov jump process onto the original memristor-based neural networks.This work is devoted to constructing a model of dynamic memristor-based neural network under semi-Markov jump process and investigating the corresponding)93(_∞synchronization control issue,and some sufficient criteria for ensuring the stability of error systems are obtained by Lyapunov functional method.Finally,an illustrative example is presented to show the validity of our presented method.3.This paper proposes a novel type of stochastic time delayed memristive neural networks(STDM-NNs),where the time delay randomly occurs in flux-controlled memristors.The newly presented STDM-NNs extend the existing dynamic memristor neural networks to the memristor time delay domain,where the stochastic time delay obeys Bernoulli distribution.In the body part of this paper,we firstly give out the mathematic model of an STDM-NN,further,the basic dynamic characteristics of such an STDM-NN are studied.More specifically,the existence and uniqueness of the equilibrium point(EP)of a reduced system of STDM-NNs are proven,and the global asymptotic stability(GAS)issue on such an EP of STDM-NNs is investigated in general.According to Lyapunov method,the conditions on GAS of STDM-NNs are established in the form of linear matrix inequalities.We demonstrate the proposed theoretical results by a numerical illustration at last.