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:Applied and Computational Analysis
SUMMARY:Generalized Sliced-Wasserstein Distances - Soheil
Kouri\, HRL Laboratories
DTSTART;TZID=Europe/London:20190423T150000
DTEND;TZID=Europe/London:20190423T160000
UID:TALK120643AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/120643
DESCRIPTION:Emerging from the optimal transportation problem a
nd due to their favorable geometric properties\, W
asserstein distances have recently attracted ample
attention from the machine learning and signal pr
ocessing communities. Wasserstein distances have b
een used in supervised\, semi-supervised\, and uns
upervised learning problems\, as well as in domain
adaptation and transfer learning. However\, the a
pplication of Wasserstein distances to high-dimens
ional probability measures is often hindered by th
eir expensive computational cost. Sliced-Wasserste
in (SW) distances\, on the other hand\, have simil
ar qualitative properties to the Wasserstein dista
nces but are significantly simpler to compute. The
simplicity of computation of this distance has mo
tivated recent work to use SW as a substitute for
the Wasserstein distances. In this presentation\,
I first review the mathematical concepts behind sl
iced Wasserstein distances. Then I introduce an en
tire class of new distances\, denoted as Generaliz
ed Sliced-Wasserstein (GSW) distances\, that exten
ds the idea of linear slicing used in SW distances
to general non-linear slicing of probability meas
ures. Finally\, I will review various applications
of SW and GSW in deep generative modeling and tra
nsfer learning.
LOCATION:MR 14
CONTACT:Matthew Thorpe
END:VEVENT
END:VCALENDAR