BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:LP relaxations for MAP inference - Adrian Weller (University of Ca
 mbridge)
DTSTART:20151022T133000Z
DTEND:20151022T150000Z
UID:TALK62015@talks.cam.ac.uk
CONTACT:Yingzhen Li
DESCRIPTION:For discrete graphical models\, we consider the combinatorial 
 optimization challenge of finding a mode configuration of variables\, that
  is a setting of all variables that has highest probability\, also known a
 s maximum a posteriori (MAP) inference. We shall provide a brief introduct
 ion to a popular method that frames the problem as an integer linear progr
 am then relaxes this to a linear program (LP) over continuous variables. F
 or computational efficiency\, the space over which this LP is performed is
  typically relaxed to an outer bound called the local polytope which enfor
 ces only pairwise consistency. We shall also discuss tighter relaxations t
 hat have recently been explored with some success\, and touch on message p
 assing methods that may be used to try to solve the problem efficiently.\n
 \nreadings:\n\n"Wainwright and Jordan 2008 Graphical models\, exponential 
 families and variational inference Section 8 (p. 195)":https://www.eecs.be
 rkeley.edu/~wainwrig/Papers/WaiJor08_FTML.pdf\n\n"David Sontag's phd thesi
 s chapter 2":http://cs.nyu.edu/~dsontag/papers/sontag_phd_thesis.pdf
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR