A New Classification Method Based on Semi-supervised Support Vector Machine
Semi-supervised learning using tag vector machine is a relatively new method of data classification and label-free.Semi-supervised support vector machines model the objective function is not smooth and fast optimization algorithm to solve the model cannot be applied.This paper presents a general three-moment method 3 times differentiable at the origin of construct quintic spline functions,construction of hinge can be used to approximate symmetry loss functions,the approximate accuracy estimation of and quintic spline functions.And on top of this,deduced five and a half times b-spline smoothing support vector machines for non-smooth a-smoothing model analyses the convergence.Broyden-Fletcher-Goldfarb-Shanno(storage)algorithm can be used in new models.Experimental results show that the new model has a better performance.
Classification algorithm Smooth Spline unction Semi-supervised support vector machine
Weijin Jiang Yao Lina Jiang Xinjun Xu Yuhui
School of Computer,National University of Defense Technology,Changsha,China;School of Computer and I Department of Computer,Hunan Radio and TV University,Changsha,China School of Computer and Information Engineering,Hunan University of Commerce,Changsha,China
国际会议
南昌
英文
1-13
2013-09-26（万方平台首次上网日期，不代表论文的发表时间）