Abstract: To improve the performance of extension neural network(ENN),a data selection method based on shadowed sets was proposed.This method was used to obtain training sample data for improving the performance of ENN.According to the characteristics of ENN,core data and boundary data were selected as training data for ENN;using shadowed-sets-based data selection method,core data and boundary data could be captured automatically.Experimental results indicate that the learning speed of the improved extension neural network(IENN) is faster than traditional ENN.Moreover,the generalization ability and the recognition accuracy are improved effectively.