一种新的支持向量回归核函数构建方法
A New Method for Constructing Kernel Function of Support Vector Regression
-
摘要: 首先，讨论了支持向量回归（support vectorreg ression，SVR）的基本原理．然后，从信息几何的角度分析了核函数的几何结构，通过共形变换（conformal transformation）构建与数据依赖（data-dependent）的核函数，使得特征空间在支持向量附近的体积元缩小，以改善SVR的机器性能．实验结果表明了方法的有效性．Abstract: The principle of support vector regression(SVR)is firstly discussed,then geometry of kernel function is analyzed from the viewpoint of information geometry,and the kernel function is contructed in data-dependent way by a conformal transformation,which reduces volume elements locally in neighborhoods of support vectors in feature space.This makes the performance of SVR improved.Simulation results show the effectiveness of the method.
下载:
