BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:GPstruct: Bayesian non-parametric structured prediction model - No
 vi Quadrianto\, Machine Learning\, University of Cambridge
DTSTART:20140117T120000Z
DTEND:20140117T130000Z
UID:TALK49110@talks.cam.ac.uk
CONTACT:Tamara Polajnar
DESCRIPTION:In this talk\, I will introduce a conceptually novel structure
 d prediction model\, GPstruct\, which is kernelised\, non-parametric\, and
  supporting Bayesian posterior inference. GPstruct can be instantiated for
  a wide range of structured objects such as linear chain\, tree\, grid\, a
 nd other general graphs. As a first proof of concept\, the model is benchm
 arked on segmentation\, chunking\, and named entity recognition of text pr
 ocessing tasks and gesture segmentation of video processing task involving
  a linear chain structure. One of practical issues of GPstruct is the memo
 ry demand which is quadratic in the number of latent variables and trainin
 g runtime that scales cubically. This prevents GPstruct from being applied
  to problems involving grid factor graphs\, which are prevalent in compute
 r vision applications. In the second part of the talk\, I will describe a 
 scaling trick based on ensemble learning\, with weak learners (predictors)
  trained on subsets of the latent variables and bootstrap data\, which can
  easily be distributed. We show experiments with 2 millions latent variabl
 es on image segmentation. Our method outperforms widely-used conditional r
 andom field models trained with pseudo-likelihood. Moreover\, it improves 
 over recent state-of-the-art marginal optimization methods in terms of pre
 dictive performance and uncertainty calibration. Finally\, it generalizes 
 well on all training set sizes.\n\nJoint work with Sebastien Bratieres\, Z
 oubin Ghahramani\, and Sebastian Nowozin.
LOCATION:FW26\, Computer Laboratory
END:VEVENT
END:VCALENDAR