Chaotic neural network with double self-feedbacks and its application

A novel model of chaotic neural network with double self-feedbacks composed of linear self-feedback and nonlinear Gauss-wavelet self-feedback is proposed in order to provide the network with both global searching ability and local characterizing ability. The single neuron with such double self-feedbacks can also exhibit complexly chaotic dynamic behaviors. Studies using the unified framework theory indicate that there exist two additional energy modifiers that can respectively provide the network with global searching ability and local characterizing ability to help the network to find globally optimal or near-optimal solutions. Although the network has complex self-feedbacks, it still can reach asymptotical stability. The simulation results on traveling salesman problems (TSP) show that the network with the double self-feedbacks has a higher probability of obtaining a global optimization solution compared with that with linear self-feedback or nonlinear Gauss-wavelet self-feedback.