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:Revisiting Huber&\;rsquo\;s M-Estimation: A Tun
ing-Free Approach - Chao Zheng (Lancaster Universi
ty)
DTSTART;TZID=Europe/London:20180130T110000
DTEND;TZID=Europe/London:20180130T120000
UID:TALK99136AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/99136
DESCRIPTION:We introduce a novel scheme to choose the scale or
robustification parameter in Huber&rsquo\;s metho
d for mean estimation and linear regression in bot
h low and high dimensional settings\, which is tun
ing-free. For robustly estimating the mean of a un
ivariate distribution\, we first consider the adap
tive Huber estimator with the robustification para
meter calibrated via the censored equation approac
h. Our theoretical results provide finite sample g
uarantees for both the estimation and calibration
parts. To further reduce the computational complex
ity\, we next develop an alternating M-estimation
procedure\, which simultaneously estimates the mea
n and variance in a sequential manner. This idea c
an be naturally extended to regression problems in
both low and high dimensions. We provide simple a
nd fast algorithms to implement this procedure und
er various scenarios and study the numerical perfo
rmance through simulations.
LOCATION:Seminar Room 2\, Newton Institute
CONTACT:INI IT
END:VEVENT
END:VCALENDAR