On the Eigenspectrum of the Gram Matrix and the Generalization Error of Kernel-PCA
Authors:
- J. Shawe-TaylorJSJ. Shawe-Taylor
- C.K.I. WilliamsCWC.K.I. Williams
- N. CristianiniNCN. Cristianini
- J. KandolaJKJ. Kandola
- 27 June 2005
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Theory
- Vol. 51 (7) , 2510-2522
Abstract
In this paper, the relationships between the eigenvalues of the m/spl times/m Gram matrix K for a kernel /spl kappa/(/spl middot/,/spl middot/) corresponding to a sample x/sub 1/,...,x/sub m/ drawn from a density p(x) and the eigenvalues of the corresponding continuous eigenproblem 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.Keywords
This publication has 21 references indexed in Scilit:
- Kernel Methods for Pattern Analysis2004
- A sharp concentration inequality with applicationsRandom Structures & Algorithms, 2000
- Random Matrix Approximation of Spectra of Integral OperatorsBernoulli, 2000
- 10.1162/153244303321897690Applied Physics Letters, 2000
- Authoritative sources in a hyperlinked environmentJournal of the ACM, 1999
- The anatomy of a large-scale hypertextual Web search engineComputer Networks and ISDN Systems, 1998
- Probability in Banach Spaces1991
- On the method of bounded differences1989
- Weighted sums of certain dependent random variablesTohoku Mathematical Journal, 1967
- Probability Inequalities for Sums of Bounded Random VariablesJournal of the American Statistical Association, 1963