Research On Optimization Of State Control Parameters In Neural Networks

The research is mainly based on the competitive neural network model.In the same model,different parameters will make the research content and method completely different,so the research content of this paper is divided into two parts.On the premise of proving the stability of each system,the parameters of the control protocol and activation function are optimized respectively.There is a progressive relationship between the two parts.After the synchronization of a fixed model,a model that jumps with state variables is constructed according to the actual situation and the concept of memristor.In this paper,an optimization algorithm——Local Search Algorithm,is proposed.According to the contents of modern cybernetics,different energy dissipation functions are given.Finally,the validity of the proposed theory is proved by mathematical experiments.The specific research contents are as follows:(1)The design of an optimal state estimator based on time-varying-delay competitive neural network is studied.By creating a new Lyapunov function,a sufficient condition for the stability of the sampling system estimator is constructed by using the techniques of LMI,free weight matrix method and so on.A numerical example is obtained by MATLAB simulation,verifying the feasibility of the theory and effectiveness.On this basis,Local Search Algorithm is used to find multiple numerical simulation examples.By calculating the given energy dissipation function,an optimal controller is obtained,which consumes the least energy.In summary,this chapter has a relatively complete understanding of neural network,competitive neural network,time-varying-delay,state control,and optimization.And other issues or their relations,fully applying the previous research results to my own research.(2)The optimal threshold selection of a competitive neural network based on memristor activation function parameters is studied.By using Filippov theory and combining the knowledge of differential inclusion,convex hull and closed set,the Lyapunov function is constructed,and the stability of the system is proved.On this basis,the system is optimized according to the energy consumption caused by the choice of memristor system and the different control matrix.In this paper,a set of constraints is proposed,and a new optimization algorithm is constructed.The control variable method is used to eliminate the influence of external interference and controller,and then the selection of T threshold of memristor is realized.MATLAB is used for mathematical experiments to prove its effectiveness.