摘要
以黎曼几何为理论依据,基于S.Amari的修正核函数思想提出了两种新的保角变换,用其对核函数进行数据依赖性改进,进一步提高支持向量机分类器泛化能力。以人工非线性分类问题为对象进行研究,仿真实验结果表明采用新保角映射可以快速显著地改善分类器泛化性能,而且能大幅度地减少支持向量的数目。
Two novel conformal transformations were proposed based on the Riemannian geometry theory and S. Amari＇s idea. And the kernel function was modified by the transformation in a data-dependent way. Our experimental results for the artificial nonlinear data set show that the generalization performance of support vector machines classifier is improved remarkably and the number of support vectors is decreased greatly.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2006年第3期570-572,共3页
Journal of System Simulation
基金
河北省科学技术研究与发展指导计划项目(02213560)
关键词
支持向量机
核函数
黎曼几何
保角变换
support vector machines
kernel function
Riemarmian geometry
conformal transformation
