A Randomized Heuristic for Kernel Parameter Selection with Large-scale Multi-class Data

Toke Jansen Hansen, Trine Julie Abrahamsen, Lars Kai Hansen

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

    1 Downloads (Pure)

    Abstract

    Over the past few years kernel methods have gained a tremendous amount of attention as existing linear algorithms can easily be extended to account for highly non-linear data in a computationally efficient manner. Unfortunately most kernels require careful tuning of intrinsic parameters to correctly model the distribution of the underlying data. For large-scale problems the multiplicative scaling in time complexity imposed by introducing free parameters in a crossvalidation setup will prove computationally infeasible, often leaving pure ad-hoc estimates as the only option. In this contribution we investigate a novel randomized approach for kernel parameter selection in large-scale multi-class data. We fit a minimum enclosing ball to the class means in Reproducing Kernel Hilbert Spaces (RKHS), and use the radius as a quality measure of the space, defined by the kernel parameter. We apply the developed algorithm to a computer vision paradigm where the objective is to recognize 72:000 objects among 1:000 classes. Compared to other distance metrics in the RKHS we find that our randomized approach provides better results together with a highly competitive time complexity.
    Original languageEnglish
    Title of host publication2011 IEEE International Workshop on Machine Learning for Signal Processing (MLSP)
    PublisherIEEE
    Publication date2011
    ISBN (Print)978-1-4577-1621-8
    ISBN (Electronic)978-1-4577-1622-5
    DOIs
    Publication statusPublished - 2011
    Event2011 IEEE International Workshop on Machine Learning for Signal Processing - Beijing, China
    Duration: 18 Sep 201121 Sep 2011
    Conference number: 21
    https://ieeexplore.ieee.org/xpl/conhome/6058570/proceeding

    Conference

    Conference2011 IEEE International Workshop on Machine Learning for Signal Processing
    Number21
    Country/TerritoryChina
    CityBeijing
    Period18/09/201121/09/2011
    Internet address

    Fingerprint

    Dive into the research topics of 'A Randomized Heuristic for Kernel Parameter Selection with Large-scale Multi-class Data'. Together they form a unique fingerprint.

    Cite this