Reduce the samples for SVM based on Euclidean distance

Propose a reduced sample algorithm for SVM based on Euclidean distance according to analysis the distribution feature of the support vectors. Firstly, this method pre-defines the Quasi-classification hypeplane. Then select the boundary samples according to the Euclidean distance from one sample point to the Quasi-classification hypeplane and get new training dataset. Because the new training dataset is the subset of the original training dataset, so this method can greatly reduce the size of the training dataset and improve the training speed. Finally, simulate with linear separation data and non-linear separation data. And the experimental results show the method is effective.