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:Group invariance and computational sufficiency - V
incent Vu (Ohio State University)
DTSTART;TZID=Europe/London:20180629T094500
DTEND;TZID=Europe/London:20180629T103000
UID:TALK107521AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/107521
DESCRIPTION:Statistical sufficiency formalizes the notion of d
ata reduction. In the decision theoretic interpret
ation\, once a model is chosen all inferences shou
ld be based on a sufficient statistic. However\,
suppose we start with a set of methods that share
a sufficient statistic rather than a specific mode
l. Is it possible to reduce the data beyond the s
tatistic and yet still be able to compute all of t
he methods? In this talk\, I'\;ll present some
progress towards a theory of "computational suffi
ciency" and show that strong reductions _can_ be m
ade for large classes of penalized M-estimators by
exploiting hidden symmetries in the underlying op
timization problems. These reductions can (1) ena
ble efficient computation and (2) reveal hidden co
nnections between seemingly disparate methods. As
a main example\, I'\;ll show how the theory pr
ovides a surprising answer to the following questi
on: "What do the Graphical Lasso\, sparse PCA\, si
ngle-linkage clustering\, and L1 penalized Ising m
odel selection all have in common?"
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:info@newton.ac.uk
END:VEVENT
END:VCALENDAR