BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Weighted Finite-state Automata - Rogier van Dalen\, Engineering De
 pt\, Cambridge University
DTSTART:20120419T130000Z
DTEND:20120419T143000Z
UID:TALK37651@talks.cam.ac.uk
CONTACT:David Duvenaud
DESCRIPTION:A weighted finite-state automaton assigns weights to strings o
 f discrete symbols. The formalism is a useful way of defining a probabilit
 y distribution over strings of potentially unbounded length. By defining t
 he automaton over two strings at the same time\, probabilistic relations b
 etween two sequences (e.g. conditional distributions) can be defined. I wi
 ll show how inference on this type of model works\, and how its parameters
  can be learnt. Another use of the formalism that I hope to discuss is to 
 define kernels between two strings.\n\nBackground material (I will not exp
 ect anyone to have read this):\n\nMehryar Mohri. \nFinite-State Transducer
 s in Language and Speech Processing. \nComputational Linguistics\, 23:2\, 
 1997.\nhttp://www.cs.nyu.edu/~mohri/postscript/cl1.ps\n\nJason Eisner.\nPa
 rameter estimation for probabilistic finite-state transducers.\nProceeding
 s of the 40th Annual Meeting of the Association for Computational Linguist
 ics\, 2002.\nhttp://www.cs.jhu.edu/~jason/papers/eisner.acl02-fst.pdf\n\nC
 orinna Cortes\, Patrick Haffner\, Mehryar Mohri.\nRational Kernels: Theory
  and Algorithms.\nJournal of Machine Learning Research (JMLR)\, vol. 5\, 2
 004.\nhttp://www.cs.nyu.edu/~mohri/postscript/jmlr.pdf
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR