Preventing Over-Fitting during Model Selection via Bayesian Regularisation of the Hyper-Parameters

Cawley, GC and Talbot, NLC (2007) Preventing Over-Fitting during Model Selection via Bayesian Regularisation of the Hyper-Parameters. Journal of Machine Learning Research, 8. pp. 841-861. ISSN 1533-7928

Full text not available from this repository. (Request a copy)

Abstract

While the model parameters of a kernel machine are typically given by the solution of a convex optimisation problem, with a single global optimum, the selection of good values for the regularisation and kernel parameters is much less straightforward. Fortunately the leave-one-out cross-validation procedure can be performed or a least approximated very efficiently in closed form for a wide variety of kernel learning methods, providing a convenient means for model selection. Leave-one-out cross-validation based estimates of performance, however, generally exhibit a relatively high variance and are therefore prone to over-fitting. In this paper, we investigate the novel use of Bayesian regularisation at the second level of inference, adding a regularisation term to the model selection criterion corresponding to a prior over the hyper-parameter values, where the additional regularisation parameters are integrated out analytically. Results obtained on a suite of thirteen real-world and synthetic benchmark data sets clearly demonstrate the benefit of this approach.

Item Type: Article
Additional Information: Special Topic on Model Selection
Faculty \ School: Faculty of Science > School of Computing Sciences
Related URLs:
Depositing User: Vishal Gautam
Date Deposited: 07 Mar 2011 14:55
Last Modified: 24 Jul 2019 15:22
URI: https://ueaeprints.uea.ac.uk/id/eprint/21597
DOI:

Actions (login required)

View Item View Item