BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY: Discriminative Methods with Structure - Simon Lacoste-Julien (Uni
 v of California at Berkeley)
DTSTART:20080311T113000Z
DTEND:20080311T123000Z
UID:TALK10986@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:Real world problems such as machine translation involve comple
 x dependencies. Generative models have provided an elegant and flexible fr
 amework to model those dependencies\, but they appear to lack robustness t
 o model misspecification compared to discriminative models for classificat
 ion. In this talk\, we present methods for leveraging the advantages of ge
 nerative models in the discriminative framework.\n\nIn the first part of t
 he talk\, we tackle the word alignment problem from natural language proce
 ssing. We formulate it as a weighted bipartite matching problem and show h
 ow to learn the weights by using a large-margin approach for structured pr
 ediction. By providing a flexible discriminative modeling framework\, we w
 ere able to cut the Alignment Error Rate in half compared to the previous 
 best performing generative models for word alignment.\n\nIn the second par
 t of the talk\, we study probabilistic topic models which have been popula
 r for modeling latent structures in text documents (as bag of words) or im
 ages (as bag of visual words). They are usually trained as generative mode
 ls with maximum likelihood estimation\, though this could be suboptimal if
  one is interested in doing classification. In contrast\, we present a dis
 criminative version of the Latent Dirichlet Allocation (LDA) model which a
 ttempts to uncover the latent structure in the documents while optimizing 
 its predictive power for the task of classification. We present results in
  the domains of document classification and scene categorization.\n\n(join
 t work with Fei Sha\, Ben Taskar\, Dan Klein and Michael I. Jordan)
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR