摘要
针对支持向量机(Support vector machines,SVMs)中大规模样本集训练速度慢且分类精度易受野点影响的问题,提出一个基于样本几何信息的支持向量机算法。其基本步骤是,首先分别求取每类样本点的壳向量和中心向量,然后将求出的壳向量作为新的训练集进行标准的SVM训练得到超平面的法向量,最后利用中心向量来更新法向量从而减少野点的影响得到最终的分类嚣。实验表明,采用这种学习策略,不仅加快了训练速度,而且在一般情况下也提高了分类精度。
Support vector machines (SVMs) need very long time when the scale of the training set is larger and the precision of classification is easily influenced by outliers. An algorithm based on geometric information of samples is proposed. Firstly, hull vectors and center vectors are obtained for each class. Then, the obtained convex hull vectors are used as the new training samples to train standard SVM and the normal vector of hyperplane is obtained. Finally, in order to weaken the influence of the outlier, center vectors are used to update the normal vector and obtain final classifier. Experiments show that the learning strategy quickens the training speed and improves the classification accuracy.
出处
《数据采集与处理》
CSCD
北大核心
2009年第3期328-334,共7页
Journal of Data Acquisition and Processing
基金
国家自然科学基金(60674108
60705004)资助项目
河南省科技厅科技计划(082102210091)资助项目
关键词
支持向量机
大规模训练集
壳向量
中心向量
support vector machines
large scale training set
hull vectors
center vectors