基于支持向量机的异常值检测的两种方法

Two methods of novelty detection based on SVM

支持向量机逐渐成为机器学习的一种方法。异常值检测是支持向量机中一种特殊的分类问题,被称为一类分类。一类分类通过核映射确定一个包含正类样本的紧致区域,以便使异常值更容易暴露出来。介绍了一些一类分类算法的基本思想。

Support Vector Machine(SVMs) have become an popular tool for machine learning task.Novelty detection is a special classification of SVM, called as one-class classification. For uncover novelty easily, one-class classification confirm a tighten region include positive kind through kernel mapping. The purpose of this article is introduction to basic idea of some one-class algorithm.