基于总间隔与代价差异算法的支持向量机  

Research on SVM Based on Total Margin and Different Cost Algorithm

作　　者：唐小平[1] 王文涛[1] 

机构地区：[1]中南民族大学计算机学院,武汉430074

出　　处：《中南民族大学学报（自然科学版）》2007年第4期84-86,112,共4页Journal of South-Central University for Nationalities：Natural Science Edition

基　　金：中南民族大学自然科学基金资助项目(YZY060006)

摘　　要：针对支持向量机在实际应用中存在的最优分类超平面的倾斜问题和推广误差界的问题,引入了总间隔与代价差异算法,对标准的支持向量机算法进行了改进.同时,针对线性和非线性两种情况,给出了详细的公式推导过程,并得出结论:基于总间隔与代价差异算法的支持向量机的性能优于标准的支持向量机.In respect of the skew of the optimal separating hyperplane and the generalization error bound that exist in the practical application of support vector machine, this paper introduces the total margin algorithm and different cost algorithm, which improves the standard support vector machine. Meanwhile, a detailed derivation of the formula in both linear and nonlinear cases is showed in the paper. And a conclusion can be made that the support vector machine based on total margin and different cost algorithm provides good performance, comparing with the standard support vector machine.

关 键 词：支持向量机 最优分类超平面 总间隔 代价差异 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]