BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:​​Time-varying dynamic Bayesian network reconstruction with in
 formation sharing​ - Frank Dondelinger 
DTSTART:20140529T140000Z
DTEND:20140529T153000Z
UID:TALK52862@talks.cam.ac.uk
CONTACT:Konstantina Palla
DESCRIPTION:​Network reconstruction is a well-studied topic\, with appli
 cations in systems biology\, pathway \nmedicine and social science among o
 ther areas. In many of the most interesting problems\, \nthe network struc
 ture does not remain static over time\, but changes as a result of environ
 mental \nfactors or outside interventions. Thus we require network reconst
 ruction techniques that can deal \nwith time-varying networks. In this tal
 k\, I present a technique based on dynamic Bayesian networks \nwith change
  points. The network structure\, parameters\, and the changepoints are all
  inferred from \nthe data using reversible-jump MCMC. In addition\, we use
  information sharing priors to leverage \nthe commonalities in network str
 ucture between adjacent time epochs. The model is applied to \ntwo real-wo
 rld examples of changing networks\; a network of muscle development genes 
 in \ndrosophila during morphogenesis\, and the IRMA synthetic biology gene
  network that can be \nturned on or off by changing the growth medium. I d
 emonstrate that the model improves on \nprevious approaches both in terms 
 of change point detection and network reconstruction.
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR