Journal article
Fast and Exact Leave-One-Out Analysis of Large-Margin Classifiers
Technometrics, Vol.64(3), pp.291-298
09/22/2021
DOI: 10.1080/00401706.2021.1967199
Abstract
Motivated by the Golub-Heath-Wahba formula for ridge regression, we first present a new leave-one-out lemma for the kernel support vector machines (SVM) and related large-margin classifiers. We then use the lemma to design a novel and efficient algorithm, named "magicsvm," for training the kernel SVM and related large-margin classifiers and computing the exact leave-one-out cross-validation error. By "magicsvm," the computational cost of leave-one-out analysis is of the same order of fitting a single SVM on the training data. We show that "magicsvm" is much faster than the state-of-the-art SVM solvers based on extensive simulations and benchmark examples. The same idea is also used to boost the computation speed of the V-fold cross-validation of the kernel classifiers.
Details
- Title: Subtitle
- Fast and Exact Leave-One-Out Analysis of Large-Margin Classifiers
- Creators
- Boxiang Wang - University of IowaHui Zou - University of Minnesota
- Resource Type
- Journal article
- Publication Details
- Technometrics, Vol.64(3), pp.291-298
- DOI
- 10.1080/00401706.2021.1967199
- ISSN
- 0040-1706
- eISSN
- 1537-2723
- Publisher
- Taylor & Francis
- Grant note
- DOI: 10.13039/100000001, name: NSF, award: 1915-842, 2015-120
- Language
- English
- Electronic publication date
- 09/22/2021
- Academic Unit
- Statistics and Actuarial Science
- Record Identifier
- 9984257740302771
Metrics
16 Record Views