The Learning of BP Neural Network Based on Evolutionary Programming

The shortcomings of the traditional methods are point out through the analysis of BP algorithm and genetic algorithm which are commonly applied to BP neural networks. A modified evolutionary programming is proposed to optimize the weight of BP neural network. The simulating results for XOR problem and four-parity problem are given through comparison among BP algorithm, traditional evolutionary algorithm and the proposed modified evolutionary algorithm. The result indicates that the modified evolutionary algorithm proposed in this paper is superior to other two algorithms.