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:Inference in Bayesian Networks using Dynamic Discr
etisation - Martin Neil\, Agena Ltd &\; David M
arquez\, Queen Mary\, University of London
DTSTART;TZID=Europe/London:20070326T140000
DTEND;TZID=Europe/London:20070326T150000
UID:TALK6674AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/6674
DESCRIPTION:We present a new approximate inference algorithm f
or use in hybrid Bayesian Networks (BNs). The algo
rithm efficiently combines dynamic discretisation
with robust propagation algorithms on junction tre
es structures. Our approach offers a significant e
xtension to Bayesian Network theory and practice b
y offering a flexible way of modelling continuous
nodes in BNs conditioned on complex configurations
of evidence and intermixed with discrete nodes as
both parents and children of continuous nodes. Ou
r algorithm is implemented in a commercial Bayesia
n Network software package\, AgenaRisk\, which all
ows model construction and testing to be carried o
ut easily.\n\nWe show how the rapid convergence of
the algorithm towards zones of high probability d
ensity\, make robust inference analysis possible e
ven in situations where\, due to the lack of infor
mation in both prior and data\, robust sampling be
comes infeasible. Generated solutions to realistic
modelling problems will be presented and compared
with solutions produced using competing technique
s such as Monte Carlo Markov Chains (MCMC)\, Fast
Fourier Transforms and others.\n
LOCATION:TCM Seminar Room\, Cavendish Laboratory\, Departme
nt of Physics
CONTACT:Oliver Stegle
END:VEVENT
END:VCALENDAR