一种基于卫向量的简化支持向量机模型

针对支持向量机(SVM)在处理大规模训练集时,训练速度和分类速度变慢的缺点,提出了一种基于卫向量的简化SVM模型.用对偶变换及求解线性规划方法提取卫向量,缩小训练集规模;在此基础上对训练得到的支持向量集,用线性相关性去除冗余支持向量,从而达到简化目的.对UCI标准数据集的实验表明:在保证不损失分类精度的前提下,该模型一定程度上改进了传统SVM,缩短了学习时间,取得了良好的效果.

A model for simplification SVM based on guard vectors

A simplification SVM model based on guard vectors is proposed for overcoming the slow speed of training and classification for large scale training set. In order to simplify SVM, the methods of dual transform and linear programming are used to distill guard vectors; based on that, the linearly dependent support vectors are eliminated from SV set. The experiments on the UCI database are done with this algorithm. Results show that in the condition of undeclined correct rate, the running time of this model is reduced and better performance than the standard SVM is achieved.