BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Classification Using Censored Functional Data - Aurore Delaigle\, 
 University of Melbourne
DTSTART:20130517T150000Z
DTEND:20130517T160000Z
UID:TALK45162@talks.cam.ac.uk
CONTACT:Richard Samworth
DESCRIPTION:We consider classification of functional data. This problem ha
 s\nreceived a lot of\nattention in the literature in the case where the cu
 rves are all observed on\nthe same\ninterval. A difficulty in applications
  is that the functional curves can be\nsupported on\nquite different inter
 vals\, in which case standard methods of analysis cannot\nbe used.\nWe are
  interested in constructing classifiers for curves of this type. More\npre
 cisely\, we\nconsider classification of functions supported on a compact i
 nterval\, in\ncases where the\ntraining sample consists of functions obser
 ved on other intervals\, which may\ndiffer\namong the training curves.\n\n
 We propose several methods\, depending on whether or not the observable\ni
 ntervals\noverlap by a significant amount. In the case where these interva
 ls differ a\nlot\, our\nprocedure involves extending the curves outside th
 e interval where they were\nobserved.\nWe suggest a new nonparametric appr
 oach for doing this.\n\nWe also introduce flexible ways of combining poten
 tial differences in shapes\nof the\ncurves from different populations\, an
 d potential differences between the\nendpoints of\nthe intervals where the
  curves from each population are observed.\nWe suggest a fully data-driven
  approach\, and illustrate the performance of\nour classifier\non some rea
 l and simulated data.\n
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0WB
END:VEVENT
END:VCALENDAR