摘要
支持向量机C-SVM及υ-SVM是目前两种最为成熟的模型,但是从形式到算法、从参数特性到参数含义,它们都相互不同,这给人们的选择带来不便。为了将这两种SVM模型统一起来,提出一种新的模型Cυ-SVM,并依据统计学习理论,研究它的解的特性。给出了新模型解的完备性条件,找出它的解及其相应的算法,并指出了υ/C既是边界支持向量个数的上界,又是支持向量总数的下界。参数设置说明,新模型完全可以实现旧模型的所有功能,而新的算法更加方便诸如文本自动分类等领域的使用。
In the development of Support Vector Machines (SVMs), different advantages have been put in C - SVM and v - SVM, which are two of the most refined models. But the forms of their formulations, as well as their al- gorithms and parameters, are different. This makes them hard to be chosen. In order to unify these two kinds of SVM formulations, this paper proposes a new kind of SVM, and, studies its solution characteristic based on the statistical learning theory. The completeness of Cv - SVM's solution is proved, and the solution and the algorithm of C10 - SVM are founded out. This paper points out that u/C is both an upper bound of the number of bounded support vectors and a lower bound of the number of support vectors. Parameter settings show that C10 - SVM can realize all the features of the old formulations. And the new algorithm is more convenient for users in many pattern recognition areas, such as automatic text classification.
出处
《计算机仿真》
CSCD
北大核心
2010年第4期188-191,共4页
Computer Simulation
关键词
模式识别
统计学习理论
支持向量机
Pattern recognition
Statistical learning theory
Support vector machines( SVM )