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:Isaac Newton Institute Seminar Series
SUMMARY:Uncertainty Quantification from a Mathematical Per
spective - Ralph Smith (North Carolina State Univ
ersity)
DTSTART;TZID=Europe/London:20180108T113000
DTEND;TZID=Europe/London:20180108T123000
UID:TALK97453AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/97453
DESCRIPTION:From both mathematical and statistical perspective
s\, the fundamental goal of Uncertainty Quantifica
tion (UQ) is to ascertain uncertainties inherent t
o parameters\, initial and boundary conditions\, e
xperimental data\, and models themselves to make p
redictions with improved and quantified accuracy.&
nbsp\; Some factors that motivate recent developme
nts in mathematical UQ analysis include the follow
ing. \; The first is the goal of quantifying u
ncertainties for models and applications whose com
plexity precludes sole reliance on sampling-based
methods. \; This includes simulation codes for
discretized partial differential equation (PDE) m
odels\, which can require hours to days to run.&nb
sp\; Secondly\, models are typically nonlinearly p
arameterized thus requiring nonlinear statistical
analysis. \; Finally\, there is often emphasis
on extrapolatory or out-of-data predictions\; e.g
.\, using time-dependent models to predict future
events. \; This requires embedding statistical
models within physical laws\, such as conservatio
n relations\, to provide the structure required fo
r extrapolatory predictions. \; Within this co
ntext\, the discussion will focus on techniques to
isolate subsets and subspaces of inputs that are
uniquely determined by data. \; We will also d
iscuss the use of stochastic collocation and Gauss
ian process techniques to construct and verify sur
rogate models\, which can be used for Bayesian inf
erence and subsequent uncertainty propagation to c
onstruct prediction intervals for statistical quan
tities of interest. \; The presentation will c
onclude with discussion pertaining to the quantifi
cation of model discrepancies in a manner that pre
serves physical structures.
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:INI IT
END:VEVENT
END:VCALENDAR