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:Machine Learning @ CUED
SUMMARY:Sparse Gaussian Process in Disease Mapping - Jarno
Vanhatalo\, Helsinki University of Technology
DTSTART;TZID=Europe/London:20080122T110000
DTEND;TZID=Europe/London:20080122T120000
UID:TALK10325AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/10325
DESCRIPTION:Disease mapping is a research area in spatial epid
emiology\, which aims to describe the overall dise
ase distribution on a map. The aim might be\, for
example\, to highlight areas of elevated or lowere
d mortality or morbidity risk. Gaussian process gi
ves a natural prior for the log risk surface\, sin
ce the spatial correlations between areas can be i
ncluded in an explicit and natural way into the mo
del via a correlation function. The drawback with
using a Gaussian process is the computational burd
en of the covariance matrix calculations and analy
tically intractable model. In this talk we conside
r sparse approximations to Gaussian process prior
to speed up the computations and approximate appro
aches for posterior inference. The sparse approxim
ations are fully and partially independent conditi
onal (FIC and PIC) and the posterior inference is
conducted with a help of Markov chain Monte Carlo
methods and expectation propagation.
LOCATION:Engineering Department\, CBL Room 438
CONTACT:Zoubin Ghahramani
END:VEVENT
END:VCALENDAR