Studies On RBF Chaotic Neural Network And It’s Applications

Neural network is a dynamic system which is highly non-linear. Chaos is uncertain phenomenon produced by determinate non-linear system. So, neural network and chaos are correlative nearly. Neural network will be stronger by introduce chaos in artificial neural netwok. Now, scholars acquire the model of chaotic neural network by introduce a self-feedback connective item with chaotic character into Hopfield neural network. The neural network with self-feedback is non-linear and with high dimension which cause the state track is ascertained difficultly and the chaotic phenomenon can be acquired. The dynamic character is very complicated in the neural network with chaotic characteristic. For instance, it is very hypersensitive to the initial condition. The network which is worked for long time will acquire absolutely different solution because of the initial condition exist tiny difference. The chaotic neural network can be used widely to information processing and optimization calculation as utilizing the complicated dynamic character.A novel transiently chaotic neural network with radial basis function based on Chen’s chaotic neural network. The new model has non-monotonous activation function composed by Sigmoid and Contrary-Multiquadric function which is a type of radial basis function. The imminent ability of this new activation function is stronger than the primary monotonous Sigmoid. Analyze the dynamics behavior of the chaotic neuron and research the effect on the dynamics character by some parameters of the Contrary Multiquadric function. Introduce classical combinatorial optimization problems-Traveling Salesman Problem (TSP). Apply this network to solving TSP with10cities, and compare the simulation results with the solution of Chen’s chaotic neural network, indicate that the chaotic neural network with radial basis function has better optimization ability.Chen’s chaotic neural network adopts linear self-feedback. The new model uses radial basis function as self-feedback item. The non-linear self-feedback behaves some new characters different from linear self-feedback. Analyze the dynamics behavior of the chaotic neuron and research the effect on the dynamics character by some parameters of the Contrary Multiquadric function. Base on the unified framework theory, figure out the sufficient condition of the energy function to be steady detailedly and acquire the academic foundation of the network to be steady. Apply this network to solving TSP with10cities and get better simulated result.Research the chaotic neural network with radial basis function disturbance and analyze the dynamics behavior of the chaotic neuron, besides seeing about the effect on the dynamics character by some parameters of the Contrary Multiquadric function. The breadth parameter is very important to the disturbance proved by analyzing the effect on the inside state. Apply this network to solving TSP with10cities and the simulated result indicates that the capability for resisting the disturbance is affected by some parameters of the network, indeed it maybe very strong in given condition.Base on the chaotic nueral network with radial basis function, present a new chaotic neuron’s dynamic system which can make the chaotic searching permanent. Analyze the time seriesof this system and prove the feasibility for keeping chaotic state permanently. Apply it to encrypt and decrypt for gray image and get satisfactory result. Analyze the key space and the statistical performance, accordingly indicate that this algorithm possesses the capabilities for resisting enumerate and resisting statistics.