BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Estimator selection: methods and calibration - Pascal Massart\, Un
 iversité de Paris-Sud\,
DTSTART:20140624T100000Z
DTEND:20140624T103000Z
UID:TALK53103@talks.cam.ac.uk
CONTACT:37296
DESCRIPTION:There is a huge literature devoted to the topic of estimator s
 election\, both from a\ntheoretical and a practical view point. Estimator 
 selection methods typically involve\ntuning parameters which drastically i
 nfluence the behaviour of the methods.\nTherefore the choice of these tuni
 ng parameters is an important issue. It is often the\ncase that performanc
 e bounds for estimator selection methods are over pessimistic\nor not prec
 ise enough to provide a valuable guide-line for practitioners. Fortunately
 \ncomputer intensive simulations can be used to understand how to calibrat
 e these\ntuning parameters in practice but nevertheless from a mathematica
 l view point\, it is\ndesirable to better understand the issue of calibrat
 ing selection methods. We shall\nsee that in a number of situations that i
 nclude penalized model selection\, it is\npossible to reduce the gap betwe
 en theory and practice by providing data-driven\nrules for choosing tuning
  parameters which are based on sharp upper and lower\nperformance bounds.
LOCATION:Centre for Mathematical Sciences\, Meeting Room 2
END:VEVENT
END:VCALENDAR