Abstract: When the number of the training set is very large, the training time and needed memory of support vector machine increase quickly. In order to solve the problem, an improved simplifying method was put forward. The samples which may be support vectors mainly distribute on both borders, so a method which fetches border samples better was used to improve simplifying rate, and decrease the number of samples by preserving border samples and deleting non-border samples. The experiment results prove that the simplifying rate and the generalization are almost unchangeable, so the method is feasible.

Key words: Support vector machine , Simplifying , Sample

摘要： 在针对大样本问题时,支持向量机所需训练时间和内存都急剧增加。为解决这一问题,提出一种改
进的支持向量机简化方案。根据能成为支持向量的样本主要分布在边界上,该方案提出改进提取边界样本的方法
提高约简率,保留边界样本并约简非边界样本来减小样本规模。经实验验证,此约简方法约简效果好,泛化性能几
乎没有损失,表明该方案有效可行。

关键词: 支持向量机 , 约简 , 样本