On the visualization of high-dimensional data
Résumé
Computing distances in high-dimensional spaces is deemed with the empty space phenomenon, which may harm distance-based algorithms for data visualization. We focus on transforming high-dimensional numeric data for their visualization using the kernel PCA 2D projection. Gaussian and p-Gaussian kernels are often advocated when confronted to such data; we propose to give some insight of their properties and behaviour in the context of a 2D projection for visualization. Also, such projections induce some artifacts, which, if not handled, should not be ignored.
Domaines
Autre [cs.OH]
Origine : Fichiers produits par l'(les) auteur(s)