BEGIN:VCALENDAR
VERSION:2.0
PRODID:icalendar-ruby
CALSCALE:GREGORIAN
METHOD:PUBLISH
BEGIN:VTIMEZONE
TZID:Europe/Vienna
BEGIN:DAYLIGHT
DTSTART:20210328T030000
TZOFFSETFROM:+0100
TZOFFSETTO:+0200
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
TZNAME:CEST
END:DAYLIGHT
BEGIN:STANDARD
DTSTART:20211031T020000
TZOFFSETFROM:+0200
TZOFFSETTO:+0100
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
TZNAME:CET
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260404T110148Z
UID:1628085600@ist.ac.at
DTSTART:20210804T160000
DTEND:20210804T180000
DESCRIPTION:Speaker: Yarin Gal\nhosted by Marco Mondelli\nAbstract: Bayesia
 n models are rooted in Bayesian statistics and easily benefit from the vas
 t literature in the field. In contrast\, deep learning lacks a solid mathe
 matical grounding. Instead\, empirical developments in deep learning are o
 ften 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 then that most modern de
 ep learning models can be cast as performing approximate inference in a Ba
 yesian setting. The implications of this are profound: we can use the rich
  Bayesian statistics literature with deep learning models\, explain away m
 any of the curiosities with ad hoc techniques\, combine results from deep 
 learning into Bayesian modelling\, and much more.In this talk I will discu
 ss a new class of methods that can capture uncertainty with a single forwa
 rd pass. In the process I will share some recent results shedding light on
  why standard softmax neural nets cannot normally capture epistemic uncert
 ainty reliably. I will then show how these insights allow us to propose mi
 nimal changes to the single softmax neural net\, with which we can now bea
 t the uncertainty predictions of a Deep Ensemble\, but with the computatio
 nal cost of a single standard neural net.
LOCATION:Zoom Meeting https://istaustria.zoom.us/j/95795195211?pwd=dG1Jb1Br
 UnZOK0czZnYvNWNjdmVmdz09  Meeting ID: 957 9519 5211 Passcode: 096012\, IST
 A
ORGANIZER:
SUMMARY:Yarin Gal: Uncertainty Estimation Using a Single Deep Deterministic
  Neural Network
URL:https://talks-calendar.ista.ac.at/events/3190
END:VEVENT
END:VCALENDAR