Research On Chaotic Neural Network Models With Disturbance

From the physiological structure of the human brain, artificial neural network is a method to study human intelligent behavior, and it has simulating function that human brain process information. Recently, it is found that chaotic theory can explain some irregular action in brain. So the research on chaotic neural network becomes a new question for discussion. The successful combination of wavelet and chaotic neural network makes chaotic neural network develop further. The research of chaotic neural network has been proven to be effective tool to solve optimization problems. So far, many wavelet chaotic neural network models have been presented. But most of them are improved on activation function and self-feedback coefficient. There is not improvement on its internal state in our country yet. This paper is concerned with the improvement of its internal state on the basis of wavelet chaotic neural network.The reason that people are interested in neural network lies on the hardware realization of neural network in the last analysis, namely constructing neural computer. When software simulation and hardware realization of neural network are studied, the instability of components will inevitably produce interference or noise, and this minimal disturbance will have an impact on chaotic neural network models. Therefore the main problem this paper study is whether the chaotic neural network models with disturbance can solve optimization problems effectively or not.The main contribution of this paper is as follows:1. The internal state of chaotic neural network model is improved by adding disturbance of sine function and wavelet function. Two new types of chaotic neural network models are presented.2. For Chen's chaotic neural network model, its energy function and the certification of its stability are presented. Accordingly, the feasibility of the network model is validated. 3. Chaotic neural network models with disturbance of sine function and wavelet function are applied in the optimization problems of the continuous function and TSP separately. The effect of different disturbance coefficient on finding the solutions of TSP with chaotic neural network is analyzed.4. Extensive numerical simulations show that the chaotic neural network with disturbance can get legal solutions and optimum solutions of TSP, if the disturbance coefficient is controlled properly. Therefore the very strong robustness and anti-interference ability are embodied in chaotic neural network.