A Neural Networks Evolving Method Based on Gene Expression Programming

An algorithm of automatic designation of neural networks using gene expression programming (GEP) is presented. The standard GEP is improved on so as to solve the problem of prematurity and slow convergence speed in optimizing neural networks. In this paper, an application of designing neural networks for XOR problem is formulated and compared with others. The results demonstrated that the proposed GEP approach is an effective method for evolving neural networks, and the performance of improved GEP is much better than that of standard GEP in that it not only has higher evolution efficiency, improving convergence rate from 45% to 81%, but has faster convergence speed with only 56% evolutionary number of standard GEP algorithm.