BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Inference Group
SUMMARY:Approximating The Hierarchical Dirichelet Distribu
 tion - Phil Cowans
DTSTART;TZID=Europe/London:20050316T150000
DTEND;TZID=Europe/London:20050316T160000
UID:TALK4303AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/4303
DESCRIPTION:This talk is very much a description of work in pr
 ogress!\n\nThe hierarchical Dirichlet distribution
  is the finite equivalent of the hierarchical Diri
 chlet process and can be applied to similar proble
 ms in language modelling and information retrieval
 . One such application is sequence prediction\, or
  deciding what comes next given a history of sampl
 es. For reasons which I will discuss\, it appears 
 reasonable to approximate the predictive distribut
 ion using an interpolated n-gram model with update
  exclusion\, but the relationship between the two 
 is not yet fully formalised. In this talk I'll des
 cribe how far I've got with this problem and the m
 ain issues still to be resolved.
LOCATION:Ryle Seminar Room\, Cavendish Laboratory
CONTACT:Phil Cowans
END:VEVENT
END:VCALENDAR