Non-logarithmic information measures, α-weighted em algorithms and speedup of learning

Y. Matsuyama

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

    3 Citations (Scopus)

    Abstract

    Starting from Renyi's α-divergence, a class of generalized EM algorithms called the α-EM algorithms of the WEM algorithms are derived. Merits of this generalization are found on speedup of learning, i.e., acceleration of convergence. Discussions include novel α-versions of logarithm, efficient scores, information matrices and the Cramer-Rao bound. The speedup is examined on Gaussian mixture learning systems.

    Original languageEnglish
    Title of host publicationIEEE International Symposium on Information Theory - Proceedings
    Pages385
    Number of pages1
    DOIs
    Publication statusPublished - 1998
    Event1998 IEEE International Symposium on Information Theory, ISIT 1998 - Cambridge, MA
    Duration: 1998 Aug 161998 Aug 21

    Other

    Other1998 IEEE International Symposium on Information Theory, ISIT 1998
    CityCambridge, MA
    Period98/8/1698/8/21

    ASJC Scopus subject areas

    • Applied Mathematics
    • Modelling and Simulation
    • Theoretical Computer Science
    • Information Systems

    Fingerprint Dive into the research topics of 'Non-logarithmic information measures, α-weighted em algorithms and speedup of learning'. Together they form a unique fingerprint.

  • Cite this

    Matsuyama, Y. (1998). Non-logarithmic information measures, α-weighted em algorithms and speedup of learning. In IEEE International Symposium on Information Theory - Proceedings (pp. 385). [708990] https://doi.org/10.1109/ISIT.1998.708990