Bayesian Estimation of Prior Variance in Source Term Determination

0445789 - ÚTIA 2016 AT eng A - Abstract
Šmídl, Václav - Hofman, Radek
Bayesian Estimation of Prior Variance in Source Term Determination.
EGU General Assembly Conference Abstracts. Vienna: EGU, 2015. s. 5563-5563.
[EGU General Assembly. 17.4.2015-22.4.2015, Vienna]
R&D Projects: GA MŠMT(CZ) 7F14287
Institutional support: RVO:67985556
Keywords : inverse modeling * Bayesian approach
Subject RIV: BB - Applied Statistics, Operational Research
http://library.utia.cas.cz/separaty/2015/AS/smidl-0445789.pdf

The classical formulation of the linear inverse problem is studied from Bayesian point of view. We show that the classical regularization is equivalent to prior covariance matrix. We formulate several parametrization of the prior covariance matrix and derive estimation algorithms for them. The advantages of teh new algorithms are demonstrated on data from teh ETEX experiment.
Permanent Link: http://hdl.handle.net/11104/0248311