摘要
针对单核函数支持向量机性能的局限性问题,提出将sigmoid核函数与高斯核函数组成一种新的混合核函数支持向量机。高斯核是典型的局部核;sigmoid核在神经网络中被证明具有良好的全局分类性能。新混合核函数结合二者的优点,其支持向量机的分类性能优于由单核函数构成的支持向量机,实验结果表明该方法的有效性。
Because the Support Vector Machine (SVM) based on single kernel function has some limitations on performance, a new SVM with mixed kernel was put forward. The new mixed kernel was constituted by sigmoid kernel and Gaussian kernel. Gaussian kemel is a typical local kernel; It can be demonstrated that sigmoid kernel derived from neural network has good global classification performance. The new mixed kernel combined the advantages of them. The experimental results proved that the classification of SVM with mixed kernel was much better than that with any single kernel on performance.
出处
《计算机应用》
CSCD
北大核心
2009年第B12期167-168,206,共3页
journal of Computer Applications
基金
国家自然科学基金资助项目(60603098)
关键词
支持向量机
混合核
sigmoid核
高斯核
全局核
局部核
Support Vector Machine( SVM)
mixed kernel
sigmoid kernel
Gaussian kernel
global kernel
local kernel