BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Conditional Counting: Approximate Multinomial Probit Regression - 
 Philipp Hennig (University of Cambridge)
DTSTART:20100524T100000Z
DTEND:20100524T110000Z
UID:TALK24619@talks.cam.ac.uk
CONTACT:Emli-Mari Nel
DESCRIPTION:There are many models for regression on a latent function from
  discrete data. An interesting case is the probit link function. It tends 
 to only be used for binary classification\, because the multi-class case i
 nvolves an intractable integral. \n\nI will present an approximate inferen
 ce algorithm for probit inference from multinomial (count) data. It define
 s an expressive prior on discrete probabilities\, allows inference directl
 y from counts (instead of individual discrete variables)\, and returns an 
 approximate Gaussian posterior. This makes it easy to combine with linear 
 Gaussian and Gaussian process models\, providing a predictive model for co
 unts over kernel Hilbert spaces. 
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Department of Physics
END:VEVENT
END:VCALENDAR