A variational approach to learning curves

Dorthe Malzahn, Manfred Opper, Thomas G. Dietterich (Editor), Suzanna Becker (Editor), Zoubin Ghahramani (Editor)

    Research output: Contribution to journalArticlepeer-review

    Abstract

    We combine the replica approach from statistical physics with a variational approach to analyze learning curves analytically. We apply the method to Gaussian process regression. As a main result we derive approximative relations between empirical error measures, the generalization error and the posterior variance.
    Original languageEnglish
    JournalAdvances in Neural Information Processing Systems
    Volume14
    Publication statusPublished - Sept 2002

    Bibliographical note

    Copyright of Massachusetts Institute of Technology Press (MIT Press). Available from Google Scholar.

    Keywords

    • Gaussian process regression
    • generalization error
    • posterior variance

    Fingerprint

    Dive into the research topics of 'A variational approach to learning curves'. Together they form a unique fingerprint.

    Cite this