BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Inference for infinite mixture models and Gaussian Process mixture
 s of experts using simple approximate MAP Inference - Alexis Boukouvalas (
 Aston University)
DTSTART:20150415T100000Z
DTEND:20150415T110000Z
UID:TALK59011@talks.cam.ac.uk
CONTACT:42017
DESCRIPTION:The Dirichlet process mixture (DPM) is a ubiquitous\, flexible
  Bayesian nonparametric statistical model. However\, full probabilistic in
 ference\nin this model is analytically intractable\, so that computational
 ly\nintensive techniques such as Gibb's sampling are required. As a result
 \,\nDPM-based methods\, which have considerable potential\, are restricted
  to\napplications in which computational resources and time for inference 
 is\nplentiful. We develop simplified yet statistically rigorous approximat
 e\nmaximum a-posteriori (MAP) inference algorithms for DPMs. This algorith
 m\nis as simple as K-means clustering\, performs in experiments as well as
 \nGibb's sampling\, while requiring only a fraction of the computational\n
 effort.  Finally\, we demonstrate how this approach can be used to\nperfor
 m inference for infinite mixtures of Gaussian Process experts.
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR