BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Generalization in Learning - Yevgeny Seldin (Hebrew University)
DTSTART:20090220T110000Z
DTEND:20090220T120000Z
UID:TALK17119@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:It is nowadays common to evaluate supervised learning algorith
 ms by their generalization abilities\, or\, in other words\, out-of-sample
  performance. In this context generalization bounds provide a solid theore
 tical ground for analysis of existing algorithms and development of new on
 es. By contrast to the supervised learning\, the situation in unsupervised
  learning is much more obscure. Up to the extent that if we are given two 
 reasonable solutions (e.g.\, two possible segmentations of an image) we ar
 e not able to give a well founded answer\, which one is better.\n\nIn my t
 alk I will show that it is possible to define and analyze generalization a
 bilities of unsupervised learning approaches similar to the way it is done
  in supervised learning. To support this approach by formal analysis I wil
 l derive PAC-Bayesian generalization bounds for density estimation. To dem
 onstrate the approach in practice I will derive and apply PAC-Bayesian gen
 eralization bounds in the context of co-clustering.
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR