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:CCIMI Seminars
SUMMARY:Deep and reliable – Uncertainty quantification usi
ng Empirical Bayesian deep neural networks. - Stef
an Franssen (TU Delft)
DTSTART;TZID=Europe/London:20220126T140000
DTEND;TZID=Europe/London:20220126T150000
UID:TALK167080AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/167080
DESCRIPTION:Deep learning is a popular tool for making inferen
ces\, as it has good performance in many practical
applications. Since data scientists use deep lear
ning\, we should give theoretical guarantees on th
e quality of the constructed estimator. For some a
pplications\, it is not enough to have theoretical
guarantees on the estimation error\, but\, in add
ition\, practitioners need to quantify uncertainty
. To quantify that\, a practitioner can construct
confidence sets. Researchers have only recently st
arted giving theoretical guarantees on the accurac
y of deep learning. The construction of confidence
statements is still an open problem. In this talk
\, I will first go over the general ideas and conc
epts\, discussing some earlier proposed methods fo
r uncertainty quantification before diving into my
contribution. I introduce a new Bayesian methodol
ogy: Empirical Bayesian deep neural networks (EBDN
N). EBDNN is the first methodology with theoretica
l guarantees: the uncertainty quantification produ
ced is valid from a frequentist point of view. Mor
eover\, EBDNN is much faster to compute than alter
native methods proposed for uncertainty quantifica
tion. Joint work with Botond Szabó.\n\n\n\n*Join Z
oom Meeting*\nhttps://maths-cam-ac-uk.zoom.us/j/99
412853967?pwd=enFwNFpZcG0zM0o2WVlRQm1LdEttdz09\n\n
Meeting ID: 994 1285 3967\nPasscode: 026816
LOCATION:Virtual (Zoom details under abstract)
CONTACT:Randolf Altmeyer
END:VEVENT
END:VCALENDAR