Fast Exact Leave-One-Out Cross-Validation of Sparse Least-Squares Support Vector Machines
Cached
Download Links
| Citations: | 28 - 0 self |
BibTeX
@MISC{Cawley_fastexact,
author = {Gavin C. Cawley and Nicola L. C. Talbot},
title = {Fast Exact Leave-One-Out Cross-Validation of Sparse Least-Squares Support Vector Machines},
year = {}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Leave-one-out cross-validation has been shown to give an almost unbiased esti-mator of the generalisation properties of statistical models, and therefore provides a sensible criterion for model selection and comparison. In this paper we show that ex-act leave-one-out cross-validation of sparse Least-Squares Support Vector Machines (LS-SVMs) can be implemented with a computational complexity of only O(ℓn 2) floating point operations, rather than the O(ℓ 2 n 2) operations of a naïve implemen-tation, where ℓ is the number of training patterns and n is the number of basis vectors. As a result, leave-one-out cross-validation becomes a practical proposition for model selection in large scale applications. For clarity the exposition concen-trates on sparse least-squares support vector machines in the context of non-linear regression, but is equally applicable in a pattern recognition setting. Key words: model selection, cross-validation, least-squares support vector machine
Keyphrases
sparse least-squares support vector machine model selection fast exact leave-one-out cross-validation leave-one-out cross-validation training pattern sensible criterion key word pattern recognition na ve implemen-tation floating point operation unbiased esti-mator computational complexity basis vector least-squares support vector machine statistical model practical proposition exposition concen-trates ex-act leave-one-out cross-validation non-linear regression large scale application generalisation property
