Due to the functionality of dynamic mapping for nonlinear complex data, BP neural network (BP-NN) as a typical neural network has increasingly been applied to a variety of applications. Although it has been successfully applied, its prominent shortcoming, such as the local optimum problem and the setting problem for the initial parameter of neural network, have not been completely eliminated. In this paper, an optimization algorithm for the architecture, weights and thresholds of neural networks using an improved gene expression programming (IGEP) was presented. The algorithm effectively combines the global search ability of GEP and the local search ability of BP-NN. To obtain a better efficiency, the basic GEP was improved by the dynamic adjustment of the fitness function, genetic operators and the number of evolutionary generations. The experimental results show that the IGEP-BP algorithm is an effective method for evolving neural network.

No abstract

The adaptive growing and pruning algorithm (AGP) has been improved, and the network pruning is based on the sigmoidal activation value of the node and all the weights of its outgoing connections. The nodes are pruned directly, but those nodes that have internal relation are not removed. The network growing is based on the idea of variance. We directly copy those nodes with high correlation. An improved AGP algorithm (IAGP) is proposed. And it improves the network performance and efficiency. The simulation results show that, compared with the AGP algorithm, the improved method (IAGP) can quickly and accurately predict traffic capacity.

No abstract