BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Online Expectation-Maximisation - Chris Anagnostopoulos\, Statisti
 cal Laboratory\, University of Cambridge
DTSTART:20100519T153000Z
DTEND:20100519T163000Z
UID:TALK24853@talks.cam.ac.uk
CONTACT:Richard Samworth
DESCRIPTION:The Expectation-Maximisation (EM) algorithm is a popular algor
 ithm for\nmaximum likelihood estimation in the presence of missing data an
 d/or latent\nvariables. In its standard form\, EM involves multiple runs t
 hrough the data\nwhich renders it impractical in online settings\, but rec
 ursive recastings\nare possible on the basis of Stochastic Approximation t
 heory. In this talk\,\nwe will focus on two such schemes: the seminal work
  on recursive EM by\nD.Titterington ("Recursive parameter estimation using
  incomplete data"\,\nJRSS-B 1984)\, as well as recent work by Cappe et al.
  ("Online EM\nfor latent data models"\, JRSS-B 2009). We will describe and
 \ncompare these two algorithms\, sketch their theoretical underpinnings\, 
 and\ndiscuss their applicability to challenging parameter estimation probl
 ems\nsuch as state-space and mixture modelling. To conclude\, we will brie
 fly\npoint to important open problems.\n
LOCATION:MR5\, CMS
END:VEVENT
END:VCALENDAR