Font Size: a A A

Robustness Analysis Of Several Types Of Neural Dynamic Systems

Posted on:2022-08-03Degree:MasterType:Thesis
Country:ChinaCandidate:M X ZhaFull Text:PDF
GTID:2518306479487444Subject:Basic mathematics
Abstract/Summary:PDF Full Text Request
Up to now,the types of artificial neural network models have become various.From the initial single-layer recursive neural network to the two-layer associative memory neural network,to the present multi-layer complex network,people's research has made the development of neural network vigorously.In the aspect of stability analysis,the stability of neural network may be damaged due to the inevitability of error,external disturbance and parameter fluctuation.The method to overcome these factors is to take the value of the system parameters in the specified interval,and unify the error and parameter float into a certain guiding function.Due to the existence of the guiding function,this kind of neural network becomes a hybrid neural network.That is to say,as time changes,the hybrid neural network can be subdivided into two parts: lead and delay.When studying its dynamic properties,finding the relation between its variable component and current term is the key to study the robustness of these hybrid neural networks.In this paper,we discuss the robustness of two kinds of hybrid neural networks and two kinds of nonlinear systems.By using matrix measure,linear approximation method and Lyapunov stability theory,the sufficient condition that the disturbed system is still exponentially stable under random perturbation is obtained.In addition,by using inequality expansion and stochastic differential equation theory,the upper bound of the interval and the upper bound of the maximum perturbation of the deviation variable are obtained.Finally,on the bidirectional associative memory neural network with two time-varying delays and random perturbations,the theoretical criterion of exponential stability of the bidirectional neural network with disturbance is derived by constructing the appropriate Lyapunov function.In general,given a self-stable neural network or nonlinear system,the antiinterference ability of the neural network or nonlinear system is demonstrated in this paper by considering the influence of external parameter deviation variables and random noise interference on the system stability.In addition,impulse effect,energy loss,network packet loss and other phenomena may occur in the neural network.In the future,the robustness of the neural network after adding these factors can be considered.
Keywords/Search Tags:Deviation argument, Hybrid neural network, Nonlinear system, Random disturbance, Robustness
PDF Full Text Request
Related items