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:Statistics
SUMMARY:Graph-Guided Banding for Covariance Estimation - J
acob Bien (Cornell University)
DTSTART;TZID=Europe/London:20151211T143000
DTEND;TZID=Europe/London:20151211T153000
UID:TALK60807AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/60807
DESCRIPTION:Reliable estimation of the covariance matrix is no
toriously difficult in high dimensions. Numerous m
ethods assume that the population covariance (or i
nverse covariance) matrix is sparse while making n
o particular structural assumptions on the desired
sparsity pattern. A highly-related\, yet compleme
ntary\, literature studies the setting in which th
e measured variables have a known ordering\, in wh
ich case a banded (or near-banded) population matr
ix is assumed. This work focuses on the broad midd
le ground that lies between the former approach of
complete neutrality to the sparsity pattern and t
he latter highly restrictive assumption of having
a known ordering. We develop a class of convex reg
ularizers that is in the spirit of banding and yet
attains sparsity structures that can be customize
d to a wide variety of applications.
LOCATION:MR12\, Centre for Mathematical Sciences\, Wilberfo
rce Road\, Cambridge.
CONTACT:Quentin Berthet
END:VEVENT
END:VCALENDAR