The Research Of The Associative Memory Function Of Chaotic Neural Network

With the development of science, artificial intelligence research has made rapid progress. To create an intelligent system which can simulate the function of human brain and has the ability of learning and thinking independently has been a common goal of many scientific research personnel. The associative memory, the most basic function of human brain, has always been the research object which is tried to been achieved by computers.Nowadays, the implementation of associative memory function based on neural network has been one of hotspots, especially the Hopfiled neural network published in1982. In the past30years, although a lot of scientists at home and abroad have improved Hopfiled neural network in order to enhance the storage capacity and fault tolerance of the network, the study of associative memory function has entered a bottleneck due to its limitations. After that, the neural network based on biological structure is put forward and it is proved that there is a chaotic phenomenon in human brain by biologists. In1990, according to the results of the bio-electrochemical experiments Ahara advanced the chaotic neural network, which has made some achievements in the last20years.This article has made further research of the associative memory function based on chaotic network on the basis of some previous study. First, the basic theories of chaotic neural network were introduced and its associative memory function was analyzed based on theory and experiments. If original chaotic neural network is in a certain range of parameters, the network neural element will enter the chaotic state, which makes an incorrect output of neurons. In order to overcome this shortcoming, this paper will propose a chaotic neural network control method, which has enhanced associative memory function proved by the experiments. At the end of this article, this method was used in the application of character recognition.