摘要
基于传统的V-SVM算法,结合C-SVM二阶范数软间隔形式,构造出一种变形算法,避免了求解其对偶问题过程中解向量大小的限制,论证其解的存在以及唯一性,对于线性不可分问题,采用核函数技术,达到求解的目的,通过编程,得到数值结果,表明这是一种有效可行的算法。
The basic V-SVM and C-SVM algorithm are commonly used in SVM（ Support Vector Machine） . There are some relationship among the number of support vectors with the parameter V in the V-SVM algorithm. In practical applications, it has no obvious effect. By solving the dual problem, each component of dual problem g solution vector is less than the inverse of total number of sample data, but each component whether is zero or not directly affects the number of support vectors, When the number of sample data are great, it is easy to produce very large errors. Based on traditional V-SVM algorithm, combined with C-SVM two order norm soft margin method, this paper proposes a new V-SVM deformational algorithm, avoids the limit of its dual problem solution, demonstrates the existence of its solution and uniqueness. Kernel function technical is used to solve the nonlinear form. By programming, the numerical results are obtained,which shows that the algorithm is effective and feasible.
出处
《成都工业学院学报》
2013年第2期22-25,共4页
Journal of Chengdu Technological University
基金
国家自然科学基金"非定常N-S方程的稳定化有限元方法"(11271273)
关键词
V—SVM
变形
解
核函数
数值试验
V-SVM
deformation
solution
kernel function
numerical experimentation
