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:Algorithmic Approaches to Statistical Estimation u
nder Structural Constraints - Ilias Diakonikolas\,
University of Edinburgh
DTSTART;TZID=Europe/London:20150206T160000
DTEND;TZID=Europe/London:20150206T170000
UID:TALK55591AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/55591
DESCRIPTION: The area of inference under structural (or shape)
constraints -- that is\, inference about a proba
bility distribution under the constraint that its
density function satisfies certain qualitative pro
perties -- is a classical topic in statistics and
machine learning. Shape restricted inference has s
een a recent surge of research activity\, in part
due to the ubiquity of structured distributions in
the natural sciences. The hope is that\, under su
ch structural constraints\, the quality of the re
sulting estimators may dramatically improve\, both
in terms of sample size and in terms of computati
onal efficiency.\n\nIn this talk\, we will describ
e a framework that yields new\, provably efficient
estimators for several natural and well-studied c
lasses of distributions. Our approach relies on a
single\, unified algorithm that provides a fairly
complete picture of the sample and computational c
omplexities for fundamental inference tasks. The f
ocus of the talk will be on density estimation (le
arning)\, but we may also discuss applications of
these ideas to hypothesis testing.
LOCATION:MR12\, Centre for Mathematical Sciences\, Wilberf
orce Road\, Cambridge
CONTACT:
END:VEVENT
END:VCALENDAR