BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Structural Learning of Dynamic Bayesian Networks - Matt Henderson 
 (University of Cambridge)
DTSTART:20120322T140000Z
DTEND:20120322T153000Z
UID:TALK37142@talks.cam.ac.uk
CONTACT:Konstantina Palla
DESCRIPTION:Dynamic Bayesian Networks allow complex systems of multiple ra
 ndom variables to be modelled as they vary over time.  \nThe conditional d
 ependencies of the variables within a time slice\, and between two adjacen
 t slices are encoded in the structure \nof the prior and transition networ
 ks respectively.  This talk introduces the problem of learning these struc
 tures automatically from \ndata.  The problem is particularly complicated 
 by the possible existence of missing data\, and  hidden variables.  A tech
 nique for \nlearning structures with hidden variables by Boyen\, Friedman 
 and Koller is presented.  The essential observation is that the omission \
 nof hidden variables in a Dynamic Bayesian Network can lead to violations 
 of the Markov assumption.\n
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR