基于填充函数法的光滑分段半监督支持向量机

摘要

构造一个光滑分段函数，以逼近梯度直推式支持向量机模型中的不可微项，得到一种新的光滑分段半监督支持向量机模型。根据新模型的非凸特性，采用无参数填充函数法对其求解，以此避免参数调节的复杂过程，减小计算量。采用 UCI数据库中不同规模的数据集，对新模型进行训练和测试，结果表明新模型可降低向量机的平均误分率，提高其分类性能。%A smooth piecewise function is constructed to approximate the non-differentiable part of the gradient transductive support vector machine,and then,a new smooth semi-supervised support vector machine model is proposed.According to the non-convex character of the new model,a new parameter free filled function method can be used to solve the smooth non-convex programming,and thus the difficulties of parameter adj ustment can be overcome,which makes the model simple and easy to implement.Several data sets of different size from the UCI database are used to train and test the new model.The results show that the error classification rate becomes more lower,and that the classification performance are improved.