Model reduction and uncertainty quantication for parameter estimation
Bulent Karasozen (Institute of Appled Mathematics and Department of
Mathematics, Middle East Technical University (METU) Ankara, Turkey) and Thomas
Carraro (Institute for Applied Mathematics, Heidelberg University Heidelberg, Germany)
Parameter estimation is widely applied in the context of model calibration using parameter fitting and estimation of the model confidence. The minisymposium focuses on
parameter estimation of PDE based models. In particular, the following main aspects
1) Model reduction
2) Uncertainty quantification
All methods for model reduction (order reduction, adaptivity etc.) that allow for a
reduction of storage and/or computational time maintaining the same response characteristics of the original system are of interest for the minisymposium. One important
aspect is the evaluation of the accuracy of the reduced model.
An additional essential aspect for parameter estimation problems is the prediction of
the confidence of the fitted parameters and/or of a quantity of interest (QI), which
depends on the fitted values. In the minisymposium techniques to estimate the variance
(and covariance) of the parameters and/or QI using a sensitivity approach are considered. Approaches as stochastic nite elements are within the scope of the minisymposium. Furthermore, methods for optimal experimental design based on deterministic
approaches are considered important in this context.