Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data

Trine Julie Abrahamsen, Lars Kai Hansen

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

    Abstract

    Variance inflation is caused by a mismatch between linear projections of test and training data when projections are estimated on training sets smaller than the dimensionality of the feature space. We demonstrate that variance inflation can lead to an increased neuroimage decoding error rate for Support Vector Machines. However, good generalization may be recovered in part by a simple renormalization procedure. We show that with proper renormalization, cross-validation based parameter optimization leads to the acceptance of more non-linearity in neuroimage classifiers than would have been obtained without renormalization.
    Original languageEnglish
    Title of host publicationMachine Learning and Interpretation in Neuroimaging : International Workshop, MLINI 2011, Held at NIPS 2011, Sierra Nevada, Spain, December 16-17, 2011, Revised Selected and Invited Contributions
    PublisherSpringer
    Publication date2012
    Pages256-263
    ISBN (Print)978-3-642-34712-2
    ISBN (Electronic)978-3-642-34713-9
    DOIs
    Publication statusPublished - 2012
    EventInternational Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI 2011) - Granada, Spain
    Duration: 16 Dec 201117 Dec 2011
    https://sites.google.com/site/mlini2011/home

    Workshop

    WorkshopInternational Workshop on Machine Learning and Interpretation in Neuroimaging (MLINI 2011)
    CountrySpain
    CityGranada
    Period16/12/201117/12/2011
    Internet address
    SeriesLecture Notes in Computer Science
    Volume7263
    ISSN0302-9743

    Keywords

    • Support Vector Machines
    • Generalizability
    • Variance inflation
    • Imbalanced data

    Fingerprint Dive into the research topics of 'Restoring the Generalizability of SVM Based Decoding in High Dimensional Neuroimage Data'. Together they form a unique fingerprint.

    Cite this