28605 articles – 22086 references  [version française]

inria-00444732, version 1

A Cramer-Rao Bound Characterization of the EM-Algorithm Mean Speed of Convergence

Cédric Herzet () a1, Valéry Ramon 2, Alexandre Renaux () b3, Luc Vandendorpe c2

IEEE Transactions on Signal Processing 56, 6 (2008) 2218-2228

Abstract: This paper deals with the mean speed of convergence of the expectation-maximization (EM) algorithm. We show that the asymptotic behavior (in terms of the number of observations) of the EM algorithm can be characterized as a function of the Cramér-Rao bounds (CRBs) associated to the so-called incomplete and complete data sets defined within the EM-algorithm framework. We particularize our result to the case of a complete data set defined as the concatenation of the observation vector and a vector of nuisance parameters, independent of the parameter of interest. In this particular case, we show that the CRB associated to the complete data set is nothing but the well-known modified CRB. Finally, we show by simulation that the proposed expression enables to properly characterize the EM-algorithm mean speed of convergence from the CRB behavior when the size of the observation set is large enough.

  • a –  INRIA
  • b –  Université de Paris-Sud Orsay
  • c –  Université Catholique de Louvain
  • 1:  TEMICS (INRIA - IRISA)
  • CNRS : UMR6074 – INRIA – Université de Rennes 1
  • 2:  Laboratoire de Télédétection et Télécommunications (TELE)
  • Université Catholique de Louvain (UCL) - Belgique
  • 3:  Laboratoire des signaux et systèmes (L2S)
  • UMR8506 CNRS – SUPELEC – Université Paris XI - Paris Sud
  • Domain : Computer Science/Signal and Image Processing
    Engineering Sciences/Signal and Image processing
  • Keywords : Convergence of numerical methods – iterative methods – maximum-likelihood estimation
 
  • inria-00444732, version 1
  • oai:hal.inria.fr:inria-00444732
  • From: 
  • Submitted on: Thursday, 7 January 2010 11:32:22
  • Updated on: Monday, 11 January 2010 12:52:59