The Analysis Of Three Classes Of Recurrent Neural Networks

Artificial Neutral Networks are nonlinear systems of coping with informations. Their characters are abilities of coping with informations like people's mind. So they are got the prosperous development. For example, handling signal, distinguishing model, automatic control, optimizing combination, making decision with assist, solving the nonlinear algebra problem, control the robot, medical treatments and so on.Delay is the phenomenon of biology and ANNs. It is the important element which can arouse to shake and instability. Because of the dynamic characteristic deciding the capability of handling informations, studying dynamic features of ANNs such as stability, attraction, and periodicity is necessary.In recent years, people find that biologic neuron is stochastic. So effects are different with the repeatedly same stimulate. In conclusion, the studying of stochastic neural networks is of importance.This paper describes one class of recurrent neural networks with variable delays, stochastic recurrent neural networks, and the existing of unique solution of second order recurrent neural network and its global stability.The first chapter is the overview of this paper. The second chapter is about preliminary knowledge.In third chapter, the new notion of co-stability is given. the co-stability of recurrent neural networks with variable delays is discussed by using linear matrix inequality, Lyapunov functional,and the easier criterion of the co-stability is obtained. And the relevant conclusions are obtained.In the forth chapter, it combines linear matrix inequality and Lyapunov functionals with stochastic analysis to study the problem of the mean square exponential stability of stochastic recurrent neural networks.In the fifth chapter, without symmetrical link matrix and differential and monotonous of input functions and output functions, the global stability of second order recurrent neural network is proved. Besides, it extends the conclusions of relevant reference.