支持向量机核函数及优化研究

基于支持向量机的推广能力,提出选择核函数的依据.若训练样本总数一定,可采取减少支持向量数的原则减小分类错误率的产生,而减少支持向量的数量取决于核函数的选择.通过增大二次项系数的绝对值提高分类精度.实验验证改进后的核函数符合Vapnik的有关分类器推广性理论,有较高的分类精度和很好的普适性.

Survey of SVM Kernel Function and Optimum

Based on the boundary of the ability of spread, bring forward the gist of choosing SVM kernel function. If the sum of example is assured, the method through reducing the number of the support vector which depends on the choice of kernel function can decrease the rate of classification inaccuracy. And increase the absolute value of the coefficient of quadratic programming to improve the precision of classification. The experimentation proved that the improved polynomial kernel function that matches the theory about the ability of spread of Vapnik brought fine classification result and good ability of general.