Why skewing works: Learning difficult boolean functions with greedy tree learners

Bernard Rosell, Lisa Hellerstein, Soumya Ray, David Page

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Abstract

    We analyze skewing, an approach that has been empirically observed to enable greedy decision tree learners to learn "difficult" Boolean functions, such as parity, in the presence of irrelevant variables. We prove that, in an idealized setting, for any function and choice of skew parameters, skewing finds relevant variables with probability 1. We present experiments exploring how different parameter choices affect the success of skewing in empirical settings. Finally, we analyze a variant of skewing called Sequential Skewing.

    Original languageEnglish (US)
    Title of host publicationICML 2005 - Proceedings of the 22nd International Conference on Machine Learning
    EditorsL. Raedt, S. Wrobel
    Pages729-736
    Number of pages8
    StatePublished - 2005
    EventICML 2005: 22nd International Conference on Machine Learning - Bonn, Germany
    Duration: Aug 7 2005Aug 11 2005

    Other

    OtherICML 2005: 22nd International Conference on Machine Learning
    CountryGermany
    CityBonn
    Period8/7/058/11/05

    Fingerprint

    Boolean functions
    Decision trees
    Experiments

    ASJC Scopus subject areas

    • Engineering(all)

    Cite this

    Rosell, B., Hellerstein, L., Ray, S., & Page, D. (2005). Why skewing works: Learning difficult boolean functions with greedy tree learners. In L. Raedt, & S. Wrobel (Eds.), ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning (pp. 729-736)

    Why skewing works : Learning difficult boolean functions with greedy tree learners. / Rosell, Bernard; Hellerstein, Lisa; Ray, Soumya; Page, David.

    ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning. ed. / L. Raedt; S. Wrobel. 2005. p. 729-736.

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

    Rosell, B, Hellerstein, L, Ray, S & Page, D 2005, Why skewing works: Learning difficult boolean functions with greedy tree learners. in L Raedt & S Wrobel (eds), ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning. pp. 729-736, ICML 2005: 22nd International Conference on Machine Learning, Bonn, Germany, 8/7/05.
    Rosell B, Hellerstein L, Ray S, Page D. Why skewing works: Learning difficult boolean functions with greedy tree learners. In Raedt L, Wrobel S, editors, ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning. 2005. p. 729-736
    Rosell, Bernard ; Hellerstein, Lisa ; Ray, Soumya ; Page, David. / Why skewing works : Learning difficult boolean functions with greedy tree learners. ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning. editor / L. Raedt ; S. Wrobel. 2005. pp. 729-736
    @inproceedings{0b76444fad9a4d9490465db8e3299ae1,
    title = "Why skewing works: Learning difficult boolean functions with greedy tree learners",
    abstract = "We analyze skewing, an approach that has been empirically observed to enable greedy decision tree learners to learn {"}difficult{"} Boolean functions, such as parity, in the presence of irrelevant variables. We prove that, in an idealized setting, for any function and choice of skew parameters, skewing finds relevant variables with probability 1. We present experiments exploring how different parameter choices affect the success of skewing in empirical settings. Finally, we analyze a variant of skewing called Sequential Skewing.",
    author = "Bernard Rosell and Lisa Hellerstein and Soumya Ray and David Page",
    year = "2005",
    language = "English (US)",
    isbn = "1595931805",
    pages = "729--736",
    editor = "L. Raedt and S. Wrobel",
    booktitle = "ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning",

    }

    TY - GEN

    T1 - Why skewing works

    T2 - Learning difficult boolean functions with greedy tree learners

    AU - Rosell, Bernard

    AU - Hellerstein, Lisa

    AU - Ray, Soumya

    AU - Page, David

    PY - 2005

    Y1 - 2005

    N2 - We analyze skewing, an approach that has been empirically observed to enable greedy decision tree learners to learn "difficult" Boolean functions, such as parity, in the presence of irrelevant variables. We prove that, in an idealized setting, for any function and choice of skew parameters, skewing finds relevant variables with probability 1. We present experiments exploring how different parameter choices affect the success of skewing in empirical settings. Finally, we analyze a variant of skewing called Sequential Skewing.

    AB - We analyze skewing, an approach that has been empirically observed to enable greedy decision tree learners to learn "difficult" Boolean functions, such as parity, in the presence of irrelevant variables. We prove that, in an idealized setting, for any function and choice of skew parameters, skewing finds relevant variables with probability 1. We present experiments exploring how different parameter choices affect the success of skewing in empirical settings. Finally, we analyze a variant of skewing called Sequential Skewing.

    UR - http://www.scopus.com/inward/record.url?scp=31844437084&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=31844437084&partnerID=8YFLogxK

    M3 - Conference contribution

    AN - SCOPUS:31844437084

    SN - 1595931805

    SN - 9781595931801

    SP - 729

    EP - 736

    BT - ICML 2005 - Proceedings of the 22nd International Conference on Machine Learning

    A2 - Raedt, L.

    A2 - Wrobel, S.

    ER -