BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Bayesian Calibration of Computer Model Ensembles - Pratola\, M (Lo
 s Alamos National Laboratory)
DTSTART:20110909T103000Z
DTEND:20110909T110000Z
UID:TALK32738@talks.cam.ac.uk
CONTACT:Mustapha Amrani
DESCRIPTION:Using field observations to calibrate complex mathematical mod
 els of a physical process allows one to obtain statistical estimates of mo
 del parameters and construct predictions of the observed process that idea
 lly incorporate all sources of uncertainty. Many of the methods in the lit
 erature use response surface approaches\, and have demonstrated success in
  many applications. However there are notable limitations\, such as when o
 ne has a small ensemble of model runs where the model outputs are high dim
 ensional. In such instances\, arriving at a response surface model that re
 asonably describes the process can be dicult\, and computational issues m
 ay also render the approach impractical.\nIn this talk we present an appro
 ach that has numerous beneifts compared to some popular methods. First\, w
 e avoid the problems associated with defining a particular regression basi
 s or covariance model by making a Gaussian assumption on the ensemble. By 
 applying Bayes theorem\, the posterior distribution of unknown calibration
  parameters and predictions of the field process can be constructed. Secon
 d\, as the approach relies on the empirical moments of the distribution\, 
 computational and stationarity issues are much reduced compared to some po
 pular alternatives. Finally\, in the situation that additional observation
 s are arriving over time\, our method can be seen as a fully Bayesian gene
 ralization of the popular Ensemble Kalman Filter.\n\n
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR