BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Bayesian dynamic modelling  of network flows - Mike West (Duke Uni
 versity)
DTSTART:20160727T090000Z
DTEND:20160727T093000Z
UID:TALK66857@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:I discuss Bayesian dynamic modelling for sequential analysis o
 f network flow count data\, linking two  classes of models which allow fas
 t\, scalable and interpretable Bayesian inference. The first class involve
 s  sets of "<i>decoupled</i>" univariate state-space models for streaming 
 count data\, able to adaptively  characterize and quantify network dynamic
 s  in real-time. These are then "<i>recoupled</i>" to define  "<i>emulatio
 n</i>" of a second class of more structured\, time-varying gravity  models
  that  allow closer  and formal dissection of network dynamics and interac
 tions among network nodes.  Evolving internet  flows on a defined network 
 of web domains in e-commerce applications provide context\, data and  exam
 ples. Bayesian model monitoring theory defines a strategy for sequential m
 odel assessment  and adaptation in cases of signaled departures of network
  flow data from  model-based predictions. <br><br>  This work builds on th
 e more general concepts and strategy of "<i>decouple/recouple</i>" for  Ba
 yesian  model emulation. That is\,  we use decoupled\, parallel and scalab
 le analyses of a set of simpler and   computationally efficient univariate
  models\, then recouple- on a sound theoretical basis- to rebuild the  lar
 ger multivariate process for more formal inferences. <br><br>Co-authors: X
 i Chen (Duke University)\,  Kaoru Irie (University of Tokyo)\, David Banks
  (Duke University)\, Jewell Thomas (MaxPoint Interactive Inc.)\, and Rob H
 aslinger (The Sync Project). <span><br><br></span><br>  <br>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR