摘要
对模式分类中的近似线性可分问题提出了一种新的近似线性支持向量机(SVM):先对近似线性分类中的训练集所形成的两类凸壳进行了相似变形,使变形后的凸壳线性可分,再用平分最近点和最大间隔法求出理想的分划超平面,然后再通过求解最大间隔法的对偶问题得到基于相似压缩的近似线性SVM。此外,还从理论和实证分析两个方面将该方法与线性可分SVM及已有的近似线性可分SVM进行了对比分析,说明了该方法的优越性与合理性。
A new SVM is presented in this paper to solve the approximately linear separable problem of pattern recognition: First,we transform the two convex hulls which are made up of the approximately separable training set to make them separable; Second,we can figure out a separating hyperplane by halving the nearest points method or maximal margin method;Then,we get the approximately linear SVM by solving the dual problem of maximal margin method.Besides,we compared the new SVM to the known SVMs through theoretical and practical analysis,and show the advantages and rationality of the new SVM.
出处
《计算机工程与应用》
CSCD
北大核心
2007年第20期173-176,共4页
Computer Engineering and Applications
基金
国家自然科学基金(the National Natural Science Foundation of China under Grant No.60373090)
航天基金(No.02 1.3 jw0504)
关键词
SVM
近似线性SVM
相似变形压缩法
最大间隔法
分划超平面
SVM
approximately linear SVM
similitude convex hulls method
maximal margin method
separating hyperplane