BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Nonparametric Bayesian Learning of Switching Dynamical Systems - E
 mily Fox (MIT)
DTSTART:20080715T130000Z
DTEND:20080715T140000Z
UID:TALK12314@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:The hierarchical Dirichlet process hidden Markov model (HDP-HM
 M) is a flexible\, nonparametric model which allows state spaces of unknow
 n size to be learned from data. In this talk\, we demonstrate some limitat
 ions of the original HDP-HMM formulation\, and propose a _sticky_ extensio
 n which allows more robust learning of smoothly varying dynamics. Using DP
  mixtures\, this formulation also allows learning of more complex\, multim
 odal emission distributions.\nAlthough the HDP-HMM and its sticky extensio
 n are very flexible time series models\, they do make a strong Markovian a
 ssumption that observations are conditionally independent given the state.
  This\nassumption is often insufficient for capturing the temporal depende
 ncies of the observations in real data. To address this issue\, we develop
  two extensions of the sticky HDP-HMM for learning\nswitching dynamical pr
 ocesses: the switching linear dynamical system (SLDS) and the switching ve
 ctor autoregressive (VAR) process.\n\nWe develop a sampling algorithm that
  combines a truncated approximation to the Dirichlet process with an effic
 ient joint sampling of the mode and state sequences. The utility and flexi
 bility of our models are demonstrated on synthetic data\, the NIST speaker
  diarization database\, sequences of dancing honey bees\,\nand the IBOVESP
 A stock index.\n\nJoint work with Erik Sudderth\, Michael Jordan\, and Ala
 n Willsky.
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR