BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Modern Deep Learning through Bayesian Eyes - Yarin Gal\, Universit
 y of Cambridge
DTSTART:20151126T130000Z
DTEND:20151126T140000Z
UID:TALK62720@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Bayesian models are rooted in Bayesian statistics\, and easily
  benefit from the vast literature in the field. In contrast\, deep learnin
 g lacks a solid mathematical grounding. Instead\, empirical developments i
 n deep learning are often justified by metaphors\, evading the unexplained
  principles at play. These two fields are perceived as fairly antipodal to
  each other in their respective communities. It is perhaps astonishing the
 n that most modern deep learning models can be cast as performing approxim
 ate inference in a Bayesian setting. The implications of this statement ar
 e profound: we can use the rich Bayesian statistics literature with deep l
 earning models\, explain away many of the curiosities with these\, combine
  results from deep learning into Bayesian modelling\, and much more. \nIn 
 this talk I will explore the new theory linking Bayesian modelling and dee
 p learning. The practical impact of the framework will be demonstrated wit
 h a range of real-world applications: from uncertainty modelling in deep l
 earning\, through training on small datasets\, to new state-of-the-art res
 ults in image processing. I will finish by surveying open problems to rese
 arch\, problems which stand at the forefront of a new and exciting field c
 ombining modern deep learning and Bayesian techniques.\n
LOCATION:Auditorium\, Microsoft Research Ltd\, 21 Station Road\, Cambridge
 \, CB1 2FB
END:VEVENT
END:VCALENDAR