1. On the eigenspectrum of the Gram matrix and the generalization error of Kernel-PCA
- Author
-
Shawe-Taylor, John, Williams, Christopher K.I., Cristianini, Nello, and Kandola, Jaz
- Subjects
Information theory -- Research ,Eigenvalues -- Research - Abstract
In this paper, the relationships between the eigen-values of the m x m Gram matrix K for a kernel K(.,.) corresponding to a sample [x.sub.1], . [x.sub.m] drawn from a density p(x) and the eigenvalues of the corresponding continuous eigen-problem is analyzed. The differences between the two spectra are bounded and a performance bound on kernel principal component analysis (PCA) is provided showing that good performance can be expected even in very-high-dimensional feature spaces provided the sample eigenvalues fall sufficiently quickly. Index Terms--Concentration bounds, Gram matrices, kernel methods, principal components analysis (PCA), Rademacher complexity, spectra of random matrices, statistical learning theory.
- Published
- 2005