BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Machine Learning @ CUED
SUMMARY:Efficient MCMC for Continuous Time Discrete State
Systems - Vinayak Rao (UCL)
DTSTART;TZID=Europe/London:20111123T110000
DTEND;TZID=Europe/London:20111123T120000
UID:TALK34594AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/34594
DESCRIPTION:A variety of phenomena are best described using dy
namical models which operate on a discrete state s
pace and in continuous time. Examples include Mark
ov jump processes\, continuous time Bayesian netwo
rks\,\nrenewal processes and other point processes
\, with applications ranging from systems biology\
, genetics\, computing networks and human-computer
interactions. Posterior computations typically in
volve approximations\nlike time discretization and
can be computationally intensive. In this talk I
will describe recent work on a class of Markov cha
in Monte Carlo methods that allow efficient comput
ations while still being exact. The core idea is t
o use an auxiliary variable Gibbs sampler based on
uniformization\, a representation of a continuous
time\ndynamical system as a Markov chain operatin
g over a discrete set of points drawn from a Poiss
on process. This is joint work with Yee Whye Teh.
If time permits\, I shall also talk about some rec
ent work on spatial point processes with David Dun
son.
LOCATION:Engineering Department\, CBL Room 438
CONTACT:Zoubin Ghahramani
END:VEVENT
END:VCALENDAR