一种新的支持向量机快速训练算法

New fast training algorithm of support vector machine

针对大规模数据集的分类中支持向量机的训练,为解决选取样本集合边界向量时需事先判定样本集合是否线性可分的问题,提出一种基于密度法的支持向量预选取方法。该方法不需要事先判定训练样本是否线性可分,具有较强的抗击噪音点和孤立点干扰的能力,并且计算简单,易于实现。实验结果证明了这种方法是有效的。

The training of support vector machine is difficult for classing a large-scale data set. Pre-extracting support vector for support vector machine training is one of the solutions to the difficulty, but the choice is very hard. A method based on density is proposed to pre-extracting support vector. This method needn＇t confirm whether the training examples are linear separable, and has strong ability of diminishing the effect of noises and outliers. This method is simple and easy to realize. Experiments show the validity of this method.