论文部分内容阅读
支持向量机在近十年成为机器学习的主要学习技术,而且已经成功应用到有监督学习问题中。Fung和Mangasarian利用支持向量机对于既有已标类别样本又有未知类别样本的训练集进行训练,方法主要是利用少量已标明类别的样本进行训练得到一个分类器的同时对于未标明类别的样本进行分类,使得间隔最大化。此优化问题中假定样本是精确的,而在现实生活中,样本通常带有统计误差。因此,考虑样本带有扰动信息的半监督两类分类问题,给出鲁棒半监督v-支持向量分类算法。该算法的参数v易于选择,而数值试验也表明该算法具有良好的稳定性和较好的分类结果。
SVM has become the main learning technology of machine learning in recent ten years and has been successfully applied to supervised learning problems. Fung and Mangasarian used SVM to train training sets with both labeled and unknown samples by using a small number of labeled samples for training to obtain a classifier and for unspecified samples Classification, to maximize separation. This optimization problem assumes that the sample is accurate, whereas in real life, the sample usually has a statistical error. Therefore, a robust semi-supervised v-support vector classification algorithm is given considering the two semi-supervised classification problems with perturbation information. The parameter v of the algorithm is easy to choose, and numerical experiments also show that the algorithm has good stability and good classification results.