Lab Home | Phone | Search
Center for Nonlinear Studies  Center for Nonlinear Studies
 Colloquia Archive 
 Postdoc Seminars Archive 
 Quantum Lunch 
 Quantum Lunch Archive 
 CMS Colloquia 
 Q-Mat Seminars 
 Q-Mat Seminars Archive 
 P/T Colloquia 
 Kac Lectures 
 Kac Fellows 
 Dist. Quant. Lecture 
 Ulam Scholar 
 CNLS Fellowship Application 
 Student Program 
 Past Visitors 
 History of CNLS 
 Maps, Directions 
 CNLS Office 
Wednesday, April 25, 2012
3:00 PM - 4:00 PM
CNLS Conference Room (TA-3, Bldg 1690)


Mixed inference algorithms for estimation and decisions in graphical models

Alexander Ihler
University of California Irvine

Significant progress has been made in developing approximate inference methods, such as the family of belief propagation algorithms, for summation tasks (computing marginal probabilities and partition functions) and MAP estimation (finding optimal configurations).However "mixed" inference tasks that include more than one such variable elimination operator are significantly more difficult. This class of problems include "marginal MAP" problems that predict a subcomponent of the full model, and decision-making problems such as Maximum Expected Utility tasks. We give a general variational framework describing mixed inference problems, in which analogues of the Bethe, tree-reweighted, and mean field approximations can be applied, resulting in new message-passing > approximations on these tasks.

Host: Misha Chertkov,, 665-8119