基于距离配对排序的支持向量预选取算法

Pre-Selection of Support Vectors Base on Distance Pairing Sorting

摘要: 支持向量机是一种基于统计学习理论的二分类方法。支持向量机最小序列化算法(SMO)是针对支持向量机的对偶问题开发的高效算法。在数据训练过程中，支持向量对于分离超平面的确定起着决定性作用，但是支持向量仅占原始样本集的一小部分，并且分布在两类数据的边界上。如果用一个包含大多数支持向量的边界向量集来替换原始样本集进行训练，这样便能在保证分类精度的前提下，缩短训练时间，提高分类速度。然而支持向量的预选取比较困难，因此为了解决该问题，本文提出了一种基于距离配对排序的支持向量预选取算法。数值实验结果表明本文的算法能够有效地预选取包含支持向量的边界向量集。

Abstract: Support Vector Machine is a binary classification method based on statistical learning theory. The Sequential minimal optimization algorithm is an efficient algorithm developed for the dual problem of SVM. In the data training process, support vectors play a decisive role in the determination of separation hyperplane. However, support vectors are only a small part of the original sample set and distributed in the boundary of two types of data. If a boundary vector set containing most support vectors is used to replace the original sample set for training, the training time can be shortened and the classification speed can be improved on the premise of guaranteeing the classi-fication accuracy. Pre-selection of support vector is difficult. In order to solve this problem, this paper proposes a support vector pre-selection algorithm based on distance pairing sort. The ex-perimental results show that the proposed algorithm can effectively pre-select the set of boundary vectors containing support vectors.