BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Structured Prediction using Linear Programming Relaxations - David
  Sontag (NYU)
DTSTART:20120703T100000Z
DTEND:20120703T110000Z
UID:TALK38791@talks.cam.ac.uk
CONTACT:Dr Daniel Roy
DESCRIPTION:Predicting structured objects such as parse trees or protein f
 olds is\noften formulated as combinatorial optimization\, where the goal i
 s to\nfind the most likely structure given the available evidence. Linear\
 nprogramming relaxations are a powerful tool for solving these\noptimizati
 on problems. Learning for structured prediction corresponds\nto inverse co
 mbinatorial optimization\, finding parameters for the\nmodel such that for
  each of the data points\, the optimal solution is\nthe desired structure.
  This talk will survey algorithms and theory\nrelating to learning for str
 uctured prediction using linear\nprogramming relaxations\, as applied to d
 ependency parsing in natural\nlanguage processing\, multi-label prediction
 \, and protein side-chain\nplacement.\n\nBased on joint work with Michael 
 Collins\, Amir Globerson\, Tommi\nJaakkola\, Terry Koo\, Ofer Meshi\, and 
 Sasha Rush.
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR