一种多项式光滑的半监督支持向量机分类算法

为了处理半监督支持向量机优化中的非凸非光滑问题,引入一个多项式光滑函数来逼近非凸的目标函数,给出的多项式函数在样本的高密度区逼近精度高,逼近精度低时出现在样本的低密度区.采用共轭梯度法求解模型.在人工数据和UCI数据库中的4个数据集上的实验结果显示,算法不仅能保证标号数据很少时的分类精度,而且不因标号数据的增多而明显提高分类性能,因此给出的分类器性能是稳定的.

Polynomial Smooth Classification Algorithm of Semi-supervised Support Vector Machines

In order to solve the nonconvex and nonsmooth problem of semi-supervised support vector classification,a polynomial smooth function was introduced in this paper which was used to approach the nonconvex objective function.The introduced polynomial function has a high approximation accuracy in high density regions of samples and poor approximation performance appear in low density regions of samples.The model was solved by the method of conjugate gradient.Experimental results on artificial and real data suppo...