BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:MCMC for doubly-intractable distributions - Iain Murray\, Gatsby C
 omputational Neuroscience Unit\, UCL
DTSTART:20060906T130000Z
DTEND:20060906T140000Z
UID:TALK5134@talks.cam.ac.uk
CONTACT:Ryan Prescott Adams
DESCRIPTION:"The model is intractable so we resort to Markov chain Monte C
 arlo" has become a standard mantra in the Bayesian statistics community. B
 ut (even for the patient) standard MCMC techniques are not a panacea — f
 or example they can not sample from the parameter-posterior of a large tre
 e-width undirected graphical model.\n\nWe recently (Proc. UAI 2006) introd
 uced a valid MCMC scheme for this problem. Our _exchange algorithm_ is sim
 pler and often performs better than the only direct competitor (Møller et
  al.\, Biometrika 93(2):451–458\, 2006). Although both require expensive
  exact sampling (Propp and Wilson\, Rand. Struct. Alg. 9(1&2):223–252 19
 96).\n\nIn this talk I give a simpler derivation of the exchange algorithm
 . I also discuss the extent to which exact sampling is required and the im
 plications for probabilistic modeling with undirected graphs.\n\nThis is w
 ork with David MacKay and Zoubin Ghahramani.
LOCATION:Ryle Seminar Room\, Cavendish Laboratory
END:VEVENT
END:VCALENDAR