Model and algorithm for kernel quantification theory Ⅳ on large-scale samples
大样本核数量化理论Ⅳ模型及算法

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.