摘要
在基于核方法的分类问题中,核函数及其参数选择皆对分类结果具有重要影响,通常基于经验选择核函数或基于多核优化方法确定核函数的权系数。针对GF空间的多项式核函数,在范数限定条件下利用多核学习方法优化多项式权系数,实现多项式核函数的优化。实验结果表明,算法优化得到的多项式核函数其分类性能优于常用的单核函数,与多核方法相当,并在分类中取得良好的效果。
In classification problem based on the kernel method,kernel function and its parameters have a significant effect on the classification performance. Generally,we select kernel function by experience and weighted coefficient of multiple kernel function which is determined by the multiple kernel optimizing method.Under the norm restricted condition,we utilize multiple kernel learning method to optimize the weighted coefficient of polynomial kernel function based on GF space,and the optimization of polynomial kernel function can be achieved. The experiments results show that the algorithm＇s classification performance based on polynomial kernel function in this paper is better than popular used single kernel function,and meet the multiple kernel method＇s match; and it perform well in the classification research.
出处
《杭州电子科技大学学报（自然科学版）》
2015年第3期77-80,共4页
Journal of Hangzhou Dianzi University：Natural Sciences
关键词
核方法
GF空间
多核学习
kernel method
GF space
multiple kernel learning
