BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Sequential Monte Carlo for graphical models: Graph decompositions 
 and Divide-and-Conquer SMC - Dr Fredrik Lindsten\, CUED
DTSTART:20141023T130000Z
DTEND:20141023T140000Z
UID:TALK55627@talks.cam.ac.uk
CONTACT:Prof. Ramji Venkataramanan
DESCRIPTION:Probabilistic graphical models (PGMs) are widely used to repre
 sent and to reason about underlying structure in high-dimensional probabil
 ity distributions. We develop a framework for using sequential Monte Carlo
  (SMC) methods for inference and learning in general PGMs. Structural info
 rmation from the PGM is used to find a collection of graph decompositions\
 , which are then used as the basis for an SMC sampler.\n\nIn the first par
 t of the talk we consider sequential decompositions\, which results in tha
 t standard SMC techniques can be used. In the second part\, we consider in
 stead an auxiliary tree decomposition. Based on this we develop a new clas
 s of SMC samplers\, Divide-and-Conquer SMC\, in which we maintain multiple
  independent populations of weighted particles. These particle populations
  are propagated\, merged\, and resampled as the method progresses up the t
 ree. We will see how this method naturally extends the standard chain-base
 d SMC framework to a method that naturally runs on trees. We illustrate em
 pirically that these approaches can outperform standard methods in terms o
 f estimation accuracy. They also open up novel parallel implementation opt
 ions and the possibility of concentrating the computational effort on the 
 most challenging parts of the problem at hand.
LOCATION:LR5\, Cambridge University Engineering Department
END:VEVENT
END:VCALENDAR