Synchronization And Controlling Of Chaotic Neural Network

A mass of researches and experiments have proved that the existences of chaotic dynamics in the neurons and neural system. Chaotic neurons and chaotic neural networks (CNNs) are simulations of neural systems and chaotic phenomena in brains. Different than classic artificial neural networks (ANNs), CNNs have abundant chaotic dynamics. Great attentions are drawn by the application of potentials of CNNs in finding globally optimal solutions, pattern recognition and intelligent information processing.The chaotic neural networks have a lot of excellent properties. However, the chaotic characteristic makes the CNNs outputs unstable, which confines applications of the CNN. Researches of the chaotic control of the CNN arise significant interests in academia. Most of the former chaotic control methods only consider the structure of the CNN. This thesis proposes the sinusoidal parameter control method combining neural dynamics of brain. The sinusoidal parameter control method uses the sinusoidal waves, the key component of brain waves, to modulate the parameter in the CNN for the purpose of chaos control. The simulation proves that the controlled CNN can converge to a desired store pattern related to. the initial state when, the frequencies of the control signal are the same as the frequencies of the brain waves of thinking activities. Thus the controlled CNN can be used in information processing.In view of chaotic dynamics, this thesis tries to add periodic pulse signals on external stimulation parameter in order to control chaos in the CNN. From the simulations, the periodic stimulation control method can make the controlled CNN converge to periodic orbits which only contains the desired store pattern and its reverse pattern. The controlled CNN can be used in information processing and pattern recognition.Synchronizations of neurons are phenomena of collective activities in brain. Researches of neural synchronization help people to understand the principles of brain activities. In this thesis, synchronization of two-dimensional coupled neurons are simulated and discussed. The neighbor threshold coupling method can make the cluster of neurons partly synchronize. The random threshold coupling method can also improve synchronization of the coupled neurons.