An approximate maximum a posteriori method with gaussian process priors

Markus Hegland*

*Corresponding author for this work

    Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

    Abstract

    The maximum a posteriori method is generalised for infinite dimensional problems and it is shown that in this case the problem can be reduced to a nonlinear variational problem. This is not a trivial generalisation as the probability density used for the finite dimensional case does not exist. It is shown how the logarithmic gradient can be used to characterise stationary points for the Gaussian process prior case. A nonconforming finite element method is suggested using sparse grids to solve the resulting variational problem.

    Original languageEnglish
    Title of host publicationContributions to Probability and Statistics
    Subtitle of host publicationApplications and Challenges - Proceedings of the International Statistics Workshop
    PublisherWorld Scientific Publishing Co. Pte Ltd
    Pages261-275
    Number of pages15
    ISBN (Print)9812703918, 9789812703910
    DOIs
    Publication statusPublished - 2006
    EventInternational Statistics Workshop on Contributions to Probability and Statistics: Applications and Challenges - Canberra, ACT, Australia
    Duration: 4 Apr 20055 Apr 2005

    Publication series

    NameContributions to Probability and Statistics: Applications and Challenges - Proceedings of the International Statistics Workshop

    Conference

    ConferenceInternational Statistics Workshop on Contributions to Probability and Statistics: Applications and Challenges
    Country/TerritoryAustralia
    CityCanberra, ACT
    Period4/04/055/04/05

    Fingerprint

    Dive into the research topics of 'An approximate maximum a posteriori method with gaussian process priors'. Together they form a unique fingerprint.

    Cite this