BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Grouping strategies for denoising - Dominique Picard
DTSTART:20120508T140000Z
DTEND:20120508T150000Z
UID:TALK37825@talks.cam.ac.uk
CONTACT:Richard Samworth
DESCRIPTION:We investigate the statistical learning approach for modeling 
 various\napplications.\nThis modeling involves  several phases which need 
 to be solved : the first\none often is an approximation step\, where we ne
 ed to translate the\nobservations into a dictionary. The choice of this di
 ctionary (wavelets\,\nneedlets\, variouslets\,...\, combinations of severa
 l bases\,...) often conceals\na significant part of investigation.\nThe se
 cond phase is the treatment of very high dimensional data (ultra-high\ndim
 ension means that the number of parameters may grow exponentially faster\n
 than the number of observations). This phase is requiring optimization\nme
 thods of different style : $l_1$ minimizers\, multi steps methods\,...\, a
 s\nwell as concentration inequalities.\nWe concentrate on two steps thresh
 olding methods and observe that making\ngroups in  the coefficients can se
 riously improve the selection and\nprediction rates. We  provide a 'boosti
 ng-grouping' strategy\, taking into\naccount this observation.
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0WB
END:VEVENT
END:VCALENDAR