BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Mathematics and Machine Learning
SUMMARY:Bayesian deep learning - Yarin Gal (Oxford)
DTSTART;TZID=Europe/London:20180313T140000
DTEND;TZID=Europe/London:20180313T150000
UID:TALK99742AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/99742
DESCRIPTION:Bayesian models are rooted in Bayesian statistics
and easily benefit from\nthe vast literature in th
e field. In contrast\, deep learning lacks a solid
\nmathematical grounding. Instead\, empirical deve
lopments in deep learning are\noften justified by
metaphors\, evading the unexplained principles at
play.\nThese two fields are perceived as fairly an
tipodal to each other in their\nrespective communi
ties. It is perhaps astonishing then that most mod
ern deep\nlearning models can be cast as performin
g approximate inference in a\nBayesian setting. Th
e implications of this are profound: we can use th
e rich\nBayesian statistics literature with deep l
earning models\, explain away many\nof the curiosi
ties with this technique\, combine results from de
ep learning\ninto Bayesian modeling\, and much mor
e.\n\nIn this talk I will review a new theory link
ing Bayesian modeling and deep\nlearning and demon
strate the practical impact of the framework with
a range\nof real-world applications. I will also e
xplore open problems for future\nresearchâ€”problems
that stand at the forefront of this new and excit
ing\nfield.\n
LOCATION:Centre for Mathematical Sciences\, MR4
CONTACT:Damon Wischik
END:VEVENT
END:VCALENDAR