Study And Application On Chaotic Neural Network And Fuzzy Chaotic Neural Network

Artificial intelligence (AI) has been developed for many years. Though a lot of achievements have been obtained, investigation of relations between brain, thinking and compute are started just now. In order to understand of information processes in the brain, it is needed to use artificial method to simulate some functions of brain. In previously decades, due to development of neurophysiology, methods have been developed to realize the information processes in brain, such as fuzzy logic (FL), artificial neural networks (ANN), chaos and so on. All these methods are new edge subjects, so researchers apt to investigate the relations between them.Each of these three subjects can reflect one aspect of information processes in brain, and the existing ANNs, such as fuzzy neural networks (FNN) and chaotic neural networks (CNN), can only behave one or two aspects of those. A new method which is based on FL, ANN and chaos is proposed in this paper. We try to use the proposed method to understand of information processes in brain.Main works of the dissertation are as follows:Firstly, introduced the model of Wang-Smith CNN, and its dynamics characteristics, such as Lyapunov exponent (LE), bifurcation and dissipative have been analyzed thoroughly. Based on this, an improved Wang-Smith CNN is proposed, whose dynamics characteristics also have been analyzed thoroughly. Simulation results affirmed the conclusions which is obtained by theory analysis, and those results also show that the improved Wang-Smith CNN is more effective than original Wang-Smith CNN in resolving combination optimal problems (COP).Due to the existing ANN can't identify uncertain nonlinear systems accurately, two models witch are based on chaotic map are proposed in this paper. One of the models is based on multi-layers feed-forward NN, and the other is based on diagonal recurrent neural networks (DRNN). Computer simulations show that the two models have better ability to identify nonlinear systems accurately especially chaotic systems.A new method to produce texture images is also proposed in this paper. The proposed method has many merits, such as easily to be realized, can produce amount of texture images and so on.Based on fuzzy number NN, the model of fuzzy chaotic neuron is proposed in this paper, whose dynamics characteristics also have been analyzed thoroughly. The method how to build fuzzy chaotic neural networks (FCNN) with fuzzychaotic neurons is also given. And we also provided method to calculate the value of weights in FCNN.