BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Bayesian nonparametric methods for non-exchangeable data - Nick Fo
 ti (Dartmouth College)
DTSTART:20130415T100000Z
DTEND:20130415T110000Z
UID:TALK44532@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:In this talk I will present a flexible framework to incorporat
 e\nside-information into Bayesian nonparametric latent variable models.  L
 atent\nvariable models have become increasingly popular in machine learnin
 g\, examples\ninclude mixture models\, topic models\, and latent feature m
 odels.  Bayesian\nnonparametric methods allow the specification of latent 
 variable models that\ncan learn an appropriate dimensionality from the obs
 erved data.\n\nThe proposed framework has nice analytic properties\, admit
 s a simple inference \nalgorithm\, and extends previous work in the field.
   I apply the framework to both a \nlatent feature model and a topic model
  (applied to the well known corpus of State \nof Union Addresses) and demo
 nstrate that incorporating side-information improves\npredictive performan
 ce relative to exchangeable versions of the models.
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR