Research On Chaotic Neural Network And Its Application

Chaotic neural network is a new science recently. Artificial neural network is to be developed for the study intelligence process of human,focus on the problem of cognizance and simulation. Since human's knowledge of real neural system is very limited,the improvement and development of artificial neural network need more detail information from neurophysiology and neuroanatomy. Recently people find that the chaos phenomenon exists in human brain,chaotic theory can explain some irregular thing in brain. So chaotic dynamics offer new chance for studying neural network,the research of chaotic neural network becomes a new task for us.Because of its complicated dynamics property,Chaotic neural network has recently attracted some attention. Unlike the gradient descent neural network,the chaotic neural networks has more complex dynamics property,and diversified attractor exist. It is just the dynamics that make it possible for the network to be a technology with abroad application foreground for information processing and optimization computation. A in-depth research is done to chaotic neural network in this paper.Considering the discord of definition of chaos synchronization and different problem in practice,this paper the definition of chaos synchronization,a class of symbolic dynamical system describing chaos map is studied.The cognitive ability of multi-layer feed-forword neural network is studied,and some methods are given that can improve convergent speed. Because of the lack of BP algorithm,two chaos learning algorithm are proposed,one is to combine the chaos mapping and the conjugate approach,the other is to give one dimension search using chaos. Both can reach optimization spot.The study of the feedback neural network,the analysis of the work principle of Hopfield neural network lead to a neural network model with chaotic character. The network's structure is similar to that of Hopfield neural network,but it has richer and more flexible dynamics than Hopfield neural network,namely,only point attractors,so that it has higher ability of searching global optimal or near-optimal solutions. By combining chaotic dynamics and converging dynamicstogether,the neural network transit gradually to Hopfield neural network is made. By introducing converging factor,the aim of controlling chaos is attained,which provides initial value of Hopfield neural network that is near to the global optimal solutions,and solve the problem of local minimum.The principle of genetic algorithm is analyzed,and the design and of genetic algorithm are studied. Through chaos optimization method embedded into the genetic algorithm. The algorithm with the combination the advantages of the genetic algorithm and chaos optimization method which need not the optimal problem function's differential and promote the ability of the genetic algorithm's locally meticulous search can be obtained with the faster convergence and the greater probability for the global solution. A chaotic sequence is inserted into the search procedure of genetic algorithm,which can overcome premature of the search by genetic algorithm and the speed of convergence is faster than standard genetic algorithm.Traveling salesman problem is combinatorial optimization problem in graph theory,it has NPC computation complexity,and lots of problem can transfer to traveling salesman problem. The computation of TSP is analyzed,then the Hopfield network method for solving TSP is given,at last we solve 10-citys traveling salesman problem and Chinese traveling salesman problem by using chaos neural network modeling.