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:Probability Theory and Statistics in High and Infi
nite Dimensions
SUMMARY:A robust and adaptive estimator for regression II
- Yannick Baraud\, Laboratoire J.A. Dieudonné\, Ni
ce
DTSTART;TZID=Europe/London:20140624T100000
DTEND;TZID=Europe/London:20140624T103000
UID:TALK53102AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/53102
DESCRIPTION:Our purpose is to present a new method for adaptiv
ely estimating a regression\nfunction when little
is known about the shape and scale of the errors.
For instance\, it\ncan cope with error distributio
ns as different as Gaussian\, Uniform\, Cauchy or
even\nwith unimodal unbounded densities. In favora
ble cases and when the true\ndistribution belongs
to the model\, the estimator is asymptotically equ
ivalent to the\nM.L.E. and\, nevertheless\, still
behaves reasonably well when the model is wrong\,\
neven in cases for which the least-squares do not
work. The assumptions that are\nneeded to get our
results are rather weak\, in particular no moment
condition is\nrequired on the errors\, and this is
why the method can adapt to both the regression\n
function\, the shape of the errors and their scale
. Moreover\, it appears that the\npractical result
s obtained by simulation are surprisingly good as
compared to more\nspecific estimators. The corresp
onding paper is available on arXiv at\nhttp://arxi
v.org/abs/1403.6057\nJoint work with Mathieu Sart.
LOCATION:Centre for Mathematical Sciences\, Meeting Room 2
CONTACT:
END:VEVENT
END:VCALENDAR