An improved entropy-based multiple kernel learning

Hideitsu Hino, Tetsuji Ogawa

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

    Abstract

    Kernel methods have been successfully used in many practical machine learning problems. However, the problem of choosing a suitable kernel is left to practitioners. One method to select the optimal kernel is to learn a linear combination of element kernels. A framework of multiple kernel learning based on conditional entropy minimization criterion (MCEM) has been proposed and it has been shown to work well for, e.g., speaker recognition tasks. In this paper, a computationally efficient implementation for MCEM, which utilizes sequential quadratic programming, is formulated. Through a comparative experiment to conventional MCEM algorithm on a speaker verification task, the proposed method is shown to offer comparable verification accuracy with considerable improvement in computational speed.

    Original languageEnglish
    Title of host publicationProceedings - International Conference on Pattern Recognition
    Pages1189-1192
    Number of pages4
    Publication statusPublished - 2012
    Event21st International Conference on Pattern Recognition, ICPR 2012 - Tsukuba
    Duration: 2012 Nov 112012 Nov 15

    Other

    Other21st International Conference on Pattern Recognition, ICPR 2012
    CityTsukuba
    Period12/11/1112/11/15

    Fingerprint

    Entropy
    Quadratic programming
    Learning systems
    Experiments

    ASJC Scopus subject areas

    • Computer Vision and Pattern Recognition

    Cite this

    Hino, H., & Ogawa, T. (2012). An improved entropy-based multiple kernel learning. In Proceedings - International Conference on Pattern Recognition (pp. 1189-1192). [6460350]

    An improved entropy-based multiple kernel learning. / Hino, Hideitsu; Ogawa, Tetsuji.

    Proceedings - International Conference on Pattern Recognition. 2012. p. 1189-1192 6460350.

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

    Hino, H & Ogawa, T 2012, An improved entropy-based multiple kernel learning. in Proceedings - International Conference on Pattern Recognition., 6460350, pp. 1189-1192, 21st International Conference on Pattern Recognition, ICPR 2012, Tsukuba, 12/11/11.
    Hino H, Ogawa T. An improved entropy-based multiple kernel learning. In Proceedings - International Conference on Pattern Recognition. 2012. p. 1189-1192. 6460350
    Hino, Hideitsu ; Ogawa, Tetsuji. / An improved entropy-based multiple kernel learning. Proceedings - International Conference on Pattern Recognition. 2012. pp. 1189-1192
    @inproceedings{a17a6c4309d4400a88d87d8d6f812807,
    title = "An improved entropy-based multiple kernel learning",
    abstract = "Kernel methods have been successfully used in many practical machine learning problems. However, the problem of choosing a suitable kernel is left to practitioners. One method to select the optimal kernel is to learn a linear combination of element kernels. A framework of multiple kernel learning based on conditional entropy minimization criterion (MCEM) has been proposed and it has been shown to work well for, e.g., speaker recognition tasks. In this paper, a computationally efficient implementation for MCEM, which utilizes sequential quadratic programming, is formulated. Through a comparative experiment to conventional MCEM algorithm on a speaker verification task, the proposed method is shown to offer comparable verification accuracy with considerable improvement in computational speed.",
    author = "Hideitsu Hino and Tetsuji Ogawa",
    year = "2012",
    language = "English",
    isbn = "9784990644109",
    pages = "1189--1192",
    booktitle = "Proceedings - International Conference on Pattern Recognition",

    }

    TY - GEN

    T1 - An improved entropy-based multiple kernel learning

    AU - Hino, Hideitsu

    AU - Ogawa, Tetsuji

    PY - 2012

    Y1 - 2012

    N2 - Kernel methods have been successfully used in many practical machine learning problems. However, the problem of choosing a suitable kernel is left to practitioners. One method to select the optimal kernel is to learn a linear combination of element kernels. A framework of multiple kernel learning based on conditional entropy minimization criterion (MCEM) has been proposed and it has been shown to work well for, e.g., speaker recognition tasks. In this paper, a computationally efficient implementation for MCEM, which utilizes sequential quadratic programming, is formulated. Through a comparative experiment to conventional MCEM algorithm on a speaker verification task, the proposed method is shown to offer comparable verification accuracy with considerable improvement in computational speed.

    AB - Kernel methods have been successfully used in many practical machine learning problems. However, the problem of choosing a suitable kernel is left to practitioners. One method to select the optimal kernel is to learn a linear combination of element kernels. A framework of multiple kernel learning based on conditional entropy minimization criterion (MCEM) has been proposed and it has been shown to work well for, e.g., speaker recognition tasks. In this paper, a computationally efficient implementation for MCEM, which utilizes sequential quadratic programming, is formulated. Through a comparative experiment to conventional MCEM algorithm on a speaker verification task, the proposed method is shown to offer comparable verification accuracy with considerable improvement in computational speed.

    UR - http://www.scopus.com/inward/record.url?scp=84874576783&partnerID=8YFLogxK

    UR - http://www.scopus.com/inward/citedby.url?scp=84874576783&partnerID=8YFLogxK

    M3 - Conference contribution

    AN - SCOPUS:84874576783

    SN - 9784990644109

    SP - 1189

    EP - 1192

    BT - Proceedings - International Conference on Pattern Recognition

    ER -