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:Statistics
SUMMARY:Bayesian inference for Markov processes with appli
cation to biochemical network dynamics - Darren Wi
lkinson\, University of Newcastle
DTSTART;TZID=Europe/London:20111014T160000
DTEND;TZID=Europe/London:20111014T170000
UID:TALK32897AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/32897
DESCRIPTION:A number of interesting statistical applications r
equire the\nestimation of parameters underlying a
nonlinear multivariate\ncontinuous time Markov pro
cess model\, using partial and noisy discrete\ntim
e observations of the system state. Bayesian infer
ence for this\nproblem is difficult due to the fac
t that the discrete time transition\ndensity of th
e Markov process is typically intractable and\ncom
putationally intensive to approximate. It turns ou
t to be possible\nto develop MCMC algorithms which
are exact\, provided that one can\nsimulate exact
realisations of the process forwards in time. Suc
h\nalgorithms\, often termed "likelihood free" or
"plug-and-play" are very\nattractive\, as they all
ow separation of the problem of model\ndevelopment
and simulation implementation from the developmen
t of\ninferential algorithms. Such techniques brea
k down in the case of\nperfect observation or high
-dimensional data\, but more efficient\nalgorithms
can be developed if one is prepared to deviate fr
om the\nlikelihood free paradigm\, at least in the
case of diffusion processes.\nThe methods will be
illustrated using examples from population\ndynam
ics and stochastic biochemical network dynamics.
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0W
B
CONTACT:Richard Samworth
END:VEVENT
END:VCALENDAR