A training data selection method is proposed for multilayer neuralnetworks (MLNNs). This method selects a small number of the trainingdata, which guarantee both gen- eralization and fast training of theMLNNs applied to pattern classification. The generalization will besatisfied using the data locate close to the boundary of the patternclasses. However, if these data are only used in the training,convergence is slow. This phenomenon is analyzed in this paper.Therefore, in the proposed method, the MLNN is first trained usingsome number of the data, which are randomly selected (Step 1).
展开▼