BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Parallel Inference and Learning with Deep Structured Distributions
  - Alexander Schwing\, University of Toronto
DTSTART:20160315T101500Z
DTEND:20160315T111500Z
UID:TALK64810@talks.cam.ac.uk
CONTACT:44515
DESCRIPTION:Many problems in real-world applications involve predicting se
 veral random variables which are statistically related. A structured model
 \, like a Markov random field\, is a great mathematical tool to encode tho
 se dependencies.\n\nWithin the first part of this talk I will discuss the 
 difficulties in finding the most likely configuration described by a struc
 tured distribution. I will present a model-parallel inference algorithm an
 d illustrate its effectiveness in jointly estimating the disparity of more
  than 12 million variables.\n\nIn the second part\, I will show how to com
 bine structured distributions with deep learning to estimate complex repre
 sentations which take into account the dependencies between the random var
 iables. To model those deep structured distributions I will present a samp
 le-parallel training algorithm and show its applicability\, among others\,
  by using a 3D scene understanding task.\n
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR