大规模数据集下支持向量机训练样本的缩减策略

Sample Reduction Strategy for Support Vector Machines with Large-Scale Data Set

大量数据下支持向量机的训练算法是SVM研究的一个重要方向和焦点。该文从分析SVM训练的问题的实质和难点出发,提出一种在训练前先求出类别质心,去除非支持向量对应的样本,从而达到缩小样本集的方法。该方法在不损失分类正确率的情况下具有更快的收敛速度,并从空间几何上解释了支持向量机的原理。仿真实验证明了该方法的可行性和有效性。

Training algorithm for large-scale support vector rnachines（SVM） is an important and active subject in the field of SVM research. After the analysis of the nature and difficulties in training SVM, a new reduction strategy is proposed in this paper for training svm with large-scale sample set. In general, class centroid is solved before training and removing the samples corresponding to non support vectors. Through this method, the number of samples is reduced before training svm. This method is fast in convergence without accurate loss and propose the explanation of SVM theory from space geometry. The results of simulation experiments show the feasibility and effectiveness of this method.