一种新的多类模式识别支撑向量机  

A NEW MULTI-CLASS SUPPORT VECTOR MACHINE FOR PATTERN RECOGNITION

作　　者：卫保国[1] 王爱民[1] 沈兰荪[1] 

机构地区：[1]北京工业大学信号与信息处理研究室,北京100022

出　　处：《模式识别与人工智能》2002年第2期178-181,共4页Pattern Recognition and Artificial Intelligence

基　　金：国家自然科学基金(No.69971004)

摘　　要：两类支撑向量机(SVM)用于模式识别具有最优的推广能力.对于常见的多类识别问题,需要构造多类SVM.本文提出一种新的基于决策树的构造方法,由此构成的多类SVM (DTSVM),与现有的方法相比,具有更快的计算速度,适用于需处理样本数较多的识别问题.Support Vector Machines (SVM) can achieve good performance when applied to small-sample pattern recognition problems. The basic support vector machine is for two-class problem. In this paper, the principle of SVM is introduced and a new support vector machine, called decision tree based SVM (DTSVM), is proposed to solve multi-class recognition problems. In the decision tree, each non-leaf node represents a SVM classifier. The decision tree is constructed by hierarchical clustering or by a priori knowledge. DTSVM is proved to be the fastest method in our experimental evaluation. It is applicable to multi-class recognition problems with a large amount of samples.

关 键 词：多类模式识别 支撑向量机 小样本学习 机器学习 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]