Robust boosting algorithm against mislabeling in multiclass problems

Takashi Takenouchi, Shinto Eguchi, Noboru Murata, Takafumi Kanamori

    Research output: Contribution to journalArticle

    13 Citations (Scopus)

    Abstract

    We discuss robustness against mislabeling in multiclass labels for classification problems and propose two algorithms of boosting, the normalized Eta-Boost.M and Eta-Boost.M, based on the Eta-divergence. Those two boosting algorithms are closely related to models of mislabeling in which the label is erroneously exchanged for others. For the two boosting algorithms, theoretical aspects supporting the robustness for mislabeling are explored.We apply the proposed two boosting methods for synthetic and real data sets to investigate the performance of thesemethods, focusing on robustness, and confirm the validity of the proposed methods.

    Original languageEnglish
    Pages (from-to)1596-1630
    Number of pages35
    JournalNeural Computation
    Volume20
    Issue number6
    DOIs
    Publication statusPublished - 2008 Jun

    Fingerprint

    Labels
    Robustness
    Divergence
    Datasets

    ASJC Scopus subject areas

    • Control and Systems Engineering
    • Artificial Intelligence
    • Neuroscience(all)

    Cite this

    Robust boosting algorithm against mislabeling in multiclass problems. / Takenouchi, Takashi; Eguchi, Shinto; Murata, Noboru; Kanamori, Takafumi.

    In: Neural Computation, Vol. 20, No. 6, 06.2008, p. 1596-1630.

    Research output: Contribution to journalArticle

    Takenouchi, Takashi ; Eguchi, Shinto ; Murata, Noboru ; Kanamori, Takafumi. / Robust boosting algorithm against mislabeling in multiclass problems. In: Neural Computation. 2008 ; Vol. 20, No. 6. pp. 1596-1630.
    @article{fd52fb66f28e4112b7aa503e743444c2,
    title = "Robust boosting algorithm against mislabeling in multiclass problems",
    abstract = "We discuss robustness against mislabeling in multiclass labels for classification problems and propose two algorithms of boosting, the normalized Eta-Boost.M and Eta-Boost.M, based on the Eta-divergence. Those two boosting algorithms are closely related to models of mislabeling in which the label is erroneously exchanged for others. For the two boosting algorithms, theoretical aspects supporting the robustness for mislabeling are explored.We apply the proposed two boosting methods for synthetic and real data sets to investigate the performance of thesemethods, focusing on robustness, and confirm the validity of the proposed methods.",
    author = "Takashi Takenouchi and Shinto Eguchi and Noboru Murata and Takafumi Kanamori",
    year = "2008",
    month = "6",
    doi = "10.1162/neco.2007.11-06-400",
    language = "English",
    volume = "20",
    pages = "1596--1630",
    journal = "Neural Computation",
    issn = "0899-7667",
    publisher = "MIT Press Journals",
    number = "6",

    }

    TY - JOUR

    T1 - Robust boosting algorithm against mislabeling in multiclass problems

    AU - Takenouchi, Takashi

    AU - Eguchi, Shinto

    AU - Murata, Noboru

    AU - Kanamori, Takafumi

    PY - 2008/6

    Y1 - 2008/6

    N2 - We discuss robustness against mislabeling in multiclass labels for classification problems and propose two algorithms of boosting, the normalized Eta-Boost.M and Eta-Boost.M, based on the Eta-divergence. Those two boosting algorithms are closely related to models of mislabeling in which the label is erroneously exchanged for others. For the two boosting algorithms, theoretical aspects supporting the robustness for mislabeling are explored.We apply the proposed two boosting methods for synthetic and real data sets to investigate the performance of thesemethods, focusing on robustness, and confirm the validity of the proposed methods.

    AB - We discuss robustness against mislabeling in multiclass labels for classification problems and propose two algorithms of boosting, the normalized Eta-Boost.M and Eta-Boost.M, based on the Eta-divergence. Those two boosting algorithms are closely related to models of mislabeling in which the label is erroneously exchanged for others. For the two boosting algorithms, theoretical aspects supporting the robustness for mislabeling are explored.We apply the proposed two boosting methods for synthetic and real data sets to investigate the performance of thesemethods, focusing on robustness, and confirm the validity of the proposed methods.

    UR - http://www.scopus.com/inward/record.url?scp=45749120676&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=45749120676&partnerID=8YFLogxK

    U2 - 10.1162/neco.2007.11-06-400

    DO - 10.1162/neco.2007.11-06-400

    M3 - Article

    C2 - 18194110

    AN - SCOPUS:45749120676

    VL - 20

    SP - 1596

    EP - 1630

    JO - Neural Computation

    JF - Neural Computation

    SN - 0899-7667

    IS - 6

    ER -