4th Leibniz MMS Days
March 20 - March 22, 2019
Kühlungsborn

Session on Applied Inverse Problems

Inverse problems often arise due to measurement limitations, as this is e.g. the case for computer tomography, when the object of interest cannot be observed directly. Similar problems are seen in material sciences, and in geosciences (non-invasive methods). In these cases regularization methods need to be developed in order to find stable reconstructions of the quantities of interest.

But regularization techniques also occur when model based simulations are used. These models often rely on certain degrees of freedom, which must then be fixed based on observed data (model fitting). If the number of model parameters is high, and/or the the data may not follow the model accurately enough (as e.g. in climate models), then regularization is useful to avoid over-fitting and to enable reliable predictions.

The mathematical tools for regularization may be deterministic, but these may also include Bayesian procedures to allow for quantification of the inherent uncertainties.

     

back to MMS Days 2019 Main Page »