BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Empirical risk minimization for heavy-tailed losses - Gábor Lugos
 i\, Pompeu Fabra University\, Barcelona
DTSTART:20140625T141500Z
DTEND:20140625T144500Z
UID:TALK53118@talks.cam.ac.uk
CONTACT:37296
DESCRIPTION:In this talk we discuss empirical risk minimization when the l
 osses are not\nnecessarily bounded and may have a distribution with heavy 
 tails. In such situations\nusual empirical averages may fail to give accur
 ate risk estimates. However\, some\nrobust mean estimators proposed in the
  literature may be used to replace empirical\nmeans. We pay particular att
 ention to empirical risk minimization based on a robust\nestimate proposed
  by Olivier Catoni. We develop performance bounds based on a\nchaining arg
 ument tailored to Catoni's mean estimator. The results are illustrated on\
 nexamples of regression function estimation and k-means clustering.\nJoint
  work with Christian Brownlees and Emilien Joly.
LOCATION:Centre for Mathematical Sciences\, Meeting Room 2
END:VEVENT
END:VCALENDAR