中文摘要：

提出了使用核空间K—means聚类算法从训练集中抽取特征边界支持向量集，在边界集上构造支持向量机的半定规划问题，由于边界集的规模比原始训练集要小，降低了半定规划支持向量机的规模，达到优化向量机的目的．在UCI数据集上的实验结果表明：所提优化方法在求解多核半定规划向量机时，比原始方法获得几倍以上的速度提升，分类精度基本不变．

英文摘要：

Kernel K-means clustering method is proposed for abstracting the border support vector data set from training data set. The semi-definite programming SVM is solved on border set. The SVM scale is reduced as the bor- der set is less than the original training data set, and the optimization of semi-definite programming is implemented. The experimental results on UCI data set show that the new SVM training time is several times less than the original one and the classification accuracy of new SVM is equals to original one.