Abstract: Artificial neural network (ANN) can be widely used in engimeering domain, but some difficulties are existed in practical application.So far there is no theory about how to optimize network stucture. In this paper a hybrid learning procedure that combines genetic algorithm (GA) and back-propagation errors algorithms is taken to research network structure designing. Applications show that optimizing the network structure by GA can improve efficiently the performance of ANN.

Key words: Artificial intelligence, Nerve network, Structural design