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:Isaac Newton Institute Seminar Series
SUMMARY:Alternating proximal gradient descent for nonconve
x regularised problems with multiconvex coupling t
erms - Mila Nikolova (CNRS (Centre national de la
recherche scientifique)\; ENS de Cachan)
DTSTART;TZID=Europe/London:20170908T090000
DTEND;TZID=Europe/London:20170908T095000
UID:TALK78451AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/78451
DESCRIPTION:Co-author: Pauline Tan
There has been an i
ncreasing interest in constrained nonconvex \;
regularized block multiconvex optimization proble
ms. We introduce an \; approach that effective
ly exploits the multiconvex structure of the coupl
ing term and enables complex application-dependent
regularization terms to be used. The proposed Alt
ernating Structure-Adapted Proximal gradient desce
nt algorithm enjoys simple well defined updates. G
lobal convergence of the algorithm to a critical p
oint is proved using the so-called Kurdyka-Lojasie
wicz \; property. What is more\, we prove that
a large class of useful objective functions obeyi
ng our assumptions are subanalytic and thus satisf
y the Kurdyka-Lojasiewicz property. Finally\, pres
ent an application of the algorithm to big-data ai
r-born sequences of images.
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:INI IT
END:VEVENT
END:VCALENDAR