Design of BP neural network based on improved differential evolution algorithm

In order to improve the convergence of differential evolution algorithm, a chaotic differential evolution algorithm is presented by using the traversal characteristic of chaotic sequence. The search process is added chaotic sequence in order to add colony diversity. BP neural network parameters are optima by adapting the proposed algorithm. BP neural network is designed by constructing the evolution colony and fitness function. The proposed algorithm is evaluated on benchmark classification problem, and the simulation result shows a good convergence performance in optimizing the weights of the BP neural network.