支持向量机新模型及其参数特性研究

Research on a New SVM and It's Parameters

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.