电子学报 ›› 2005, Vol. 33 ›› Issue (7): 1200-1203.

• 论文 • 上一篇    下一篇

带拒识能力的双层支持向量模型分类器

胡正平, 张晔   

  1. 哈尔滨工业大学信息工程系,黑龙江哈尔滨 150001
  • 收稿日期:2004-08-18 修回日期:2005-03-09 出版日期:2005-07-25 发布日期:2005-07-25

A Two-Layer Support Vector Classifier with Rejection Feature

HU Zheng-ping, ZHANG Ye   

  1. Department of Information Engineering,Harbin Institute of Technology,Harbin,Heilongjiang 150001,China
  • Received:2004-08-18 Revised:2005-03-09 Online:2005-07-25 Published:2005-07-25

摘要: 本文构造了一种带拒识能力的双层支持向量模型分类器.在训练学习过程中,首先对各类样本特征空间求取最小的包含球形边界,得到各类样本的球形支持向量域表示.这样对于输入的非目标样本即可利用各类的支持向量域进行拒识或接受处理;然后针对接受的样本再利用基于超平面分割的SVM训练器进行分类判决.无论是在第一层求取边界的优化问题中,还是在第二层的分类超平面优化过程中,都采用相乘性更新迭代规则直接求解,优化速度与最小二乘支持向量机(LS-SVM)相当.仿真实验表明本文提出的通过引入拒绝层和判决层的新支持向量模型策略是合理可行的,在实际模式识别领域具有广阔的应用前景.

关键词: 支持向量分类器, 核函数, 支持向量域描述, 拒识性能

Abstract: A two-layer support vector classifier model with rejection feature is proposed in this paper.Firstly the sphere support vectors of each class to describe the distribution of the sample were obtained by searching all the sphere boundaries containing the samples of each class.Then the input pattern of no-object classes could be rejected by the first support vector domain description (SVDD).If a pattern is accepted by the first SVDD,the second layer of support vector classifier (SVC) with maximum margin between two classes will be used for classification.In addition,Instead of the traditional quadratic programming,multiplicative iterative updates rule is used to solve the optimizing problems in SVDD of the first layer and the SVC in second layer.Compared to the tradition algorithm of the support vector machine,the new method improves greatly the computational speed of optimization.Experimental results demonstrate that the method of two-Layer support vector classifier with Rejection Feature is feasible and it could be applicable in many real pattern recognition fields.

Key words: support vector classifier, kernel function, support vector data description, rejection performance

中图分类号: