BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Variational Approximations of Bayesian Inference
DTSTART:20111118T101500
DTSTAMP:20260407T064131Z
UID:1968f45330fe69be5584f4808c3b59adec466a444d90d2833d0f82d3
CATEGORIES:Conferences - Seminars
DESCRIPTION:Pr. M. Seeger\, Laboratory for Probabilistic Machine Learning\
 , EPFL.\nBayesian decision-making is driven by queries to the posterior di
 stribution\, obtained by conditioning a probabilistic model and prior know
 ledge on acquired data. This process is computationally challenging in gen
 eral and remains out of reach for large-scale coupled models of images or 
 image sequences. We give an overview of recent variational techniques\, wi
 th which the Bayesian computational challenge can be met approximately. In
 tractable posterior integrations are relaxed to variational optimization p
 roblems. Using novel decoupling techniques and double loop algorithms\, th
 ese problems can be mapped to standard large scale optimization primitives
  such as penalized least squares estimation and Gaussian variance computat
 ions. We present results of our methodology for the problem of optimizatio
 n magnetic resonance imaging acquisitions by Bayesian sequential optimal d
 esign.
LOCATION:ME C2 405
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR