Optimal Cross-Validation Split Ratio: Experimental Investigation

Cyril Goutte, Jan Larsen

    Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review


    Cross-validation is a common method for assessing the generalisation ability of a model in order to tune a regularisation parameter or otherhyper-parameters of a learning process. The use of cross-validation requires to set yet an additional parameter, the split rati. While a few texts haveinvestigated theoretically the asymptotic setting of this ratio, no consensus has emerged. In this contribution, we investigate the sensitivity and optimalsetting of the split ratio on a particular model, a non-parametric kernel estimator with adaptive metric.
    Original languageEnglish
    Title of host publicationProceedings of ICANN´98
    Place of PublicationLondon
    Publication date1998
    Publication statusPublished - 1998
    EventICANN´98, Proceedings of the 8th Int.Conf. on Artificial Neural Networks - Skoevde, Sweden
    Duration: 1 Jan 1998 → …


    ConferenceICANN´98, Proceedings of the 8th Int.Conf. on Artificial Neural Networks
    CitySkoevde, Sweden
    Period01/01/1998 → …

    Cite this