EMアルゴリズムの新展開 : 変分ベイズ法  [in Japanese] New Development of the EM algorithm : Variational Bayes  [in Japanese]

Access this Article

  • CiNii Fulltext PDF

    Subscription

Search this Article

Abstract

本稿では, 近年, 実践的ベイズ学習法として注目されつつある変分ベイズ法について解説する.まず, 最尤推定値を求める数値解法であるEM法および一般化EM(GEM)法について説明する.次いで, VB法の基本原理がGEMで用いられている変分近似法をベイズ拡張したものと見なせることを示す.すなわち, VB法がEM法からどのように発展して誕生したのかを系統的に解説する.

This report provides a tutorial on Variational Bayes (VB), a practical framework for Bayesian computations. First, I review the EM method which is a general procedure for obtaining maximum likelihood estimates and also explain the generalized EM (GEM) method. Then, the basic principle of the VB method can be interpreted as Bayesian extension of variational approximation used in the GEM method. Namely, I explain how the EM method has develped into the VB method.

Journal

IEICE technical report. Neurocomputing   [List of Volumes]

IEICE technical report. Neurocomputing 101(616), 23-30, 2002-01-22  [Table of Contents]

The Institute of Electronics, Information and Communication Engineers

References:  10

Cited by:  2

Codes

  • NII Article ID (NAID) :
    110003233998
  • NII NACSIS-CAT ID (NCID) :
    AN10091178
  • Text Lang :
    JPN
  • Article Type :
    Journal Article
  • ISSN :
    09135685
  • NDL Article ID :
    6074164
  • NDL Source Classification :
    ZN33(科学技術--電気工学・電気機械工業--電子工学・電気通信)
  • NDL Call No. :
    Z16-940
  • Databases :
    CJP  CJPref  NDL  NII-ELS