Improved simulated annealing mechanics in transiently chaotic neural network

The paper analyses the dynamic characteristics of transiently chaotic neural networks (TCNN), finding that they quite sensitively depend on the value of the self-feedback connection weights, and researches the annealing function that intensively influences the veracity and search speed of the TCNN model. Improved simulated annealing mechanics are proposed for the value of the self-feedback connection weights that can accelerate the search speed and guarantee the accuracy of the optimal arithmetic. To demonstrate the validity of these mechanics, two examples of function optimization problems are given.