BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Bayesian Nonparametric Model for Power Disaggregation - Isabel Val
 era (University Carlos III in Madrid)
DTSTART:20140217T110000Z
DTEND:20140217T120000Z
UID:TALK50807@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:We propose an *infinite factorial unbounded-state hidden Marko
 v model* (HMM) through the construction of a Bayesian nonparametric (BNP) 
 prior over integer-valued matrices (in which each column represents a Mark
 ov chain) with the property of presenting an infinite number of columns wi
 th an unbounded number of states\, namely\, IFUHMM. First\, we extend the 
 existent infinite factorial binary-state HMM to allow for any number of st
 ates\, and derive two Markov chain Monte Carlo (MCMC) and a variational in
 ference algorithms. Then\, we modify this model to allow an unbounded numb
 er of states and derive a new inference algorithm based on MCMC methods th
 at properly deals with the trade-off between the unbounded number of state
 s and chains. Finally\, we apply both the infinite factorial (nonbinary) H
 MM and the IFUHMM to the *power disaggregation problem* and show their abi
 lity to provide more interpretable representations of the data structure t
 han the existing BNP approaches for HMMs.
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR