大样本核数量化理论Ⅳ模型及算法  

Model and algorithm for kernel quantification theory Ⅳ on large-scale samples

作　　者：陈永良[1] 李学斌[1] 

机构地区：[1]吉林大学综合信息矿产预测研究所,长春130026

出　　处：《地球物理学进展》2010年第1期310-315,共6页Progress in Geophysics

基　　金：国家自然科学基金项目(40872193)资助

摘　　要：把核函数理论与数量化理论IV模型有机结合,提出了核数量化理论IV模型,并以高阶对称矩阵端部特征对求解的Lanczos算法为基础设计了大样本核数量化理论IV模型的算法框架.把核数量化理论IV模型应用于高光谱图像的降维实验研究,研究结果表明,合理选择核函数模型及参数,核数量化理论IV模型能够在低维标度空间中表征原始数据的族群信息,从而取得满意的分类效果.核数量化理论IV模型为大样本地学观测数据的分析处理提供了一种新的理论工具.In this paper, a kernel quantification theory IV is proposed through organical combining kernel function theory and quantification theory IV. The algorithm framework for the new model with large scale-sampling data is established on the basis of Lanczos algorithm which is an iterative method for finding the eigenpairs of a square matrix. We conduct an experiment on applying the kernel quantification theory IV to the dimension reduction of hyperspectral remote sensing images, The results show that the kernel quantification theory IV can express the clustering information of the original data in low-dimensional scaling space and get a satisfying clustering result if the kernel function and its parameters are properly selected. The kernel quantification theory IV provides an effective theoretical tool for processing large-scale sampling data in geosciences.

关 键 词：核函数 数量化理论Ⅳ LANCZOS算法 高光谱遥感图像 

分 类 号：P628[天文地球—地质矿产勘探] P631[天文地球—地质学]