An improved adaptive transiently chaotic neural-network

In this paper, we propose an improved adaptive transiently chaotic neural network. It can control the effect of energy function on neuro-dynamics during searching process of the transiently chaotic neural network (TCNN) to find global minimum efficiently. In TCNN, there exists a parameter that represents energy function´s effect. Not like existing methods to increase the parameter monotonously, our new method tries to adjust the parameter according to the change of the energy function during the neural network search process. Simulation results show that our method can converge to a stable equilibrium point fast while keeping the rate of global minima, and its performance is better than currently existing methods.