Per Christian Hansen
Informatics and Mathematical Modelling
Technical University of Denmark
pch@imm.dtu.dk
Inverse problems arise in geophysics, tomography, image deblurring and many other areas where the goal is to compute interior or hidden information from exterior data. A common feature of these ill-posed problems is that the solution is extremely sensitive to perturbations, and hence some form of stabilization or regularization is needed in order to compute a meaningful solution.
This talk will present a survey of numerical methods and paradigms suited for large-scale inverse problems. In particular we will discuss the regularizing properties of Krylov subspace methods, and we will demonstrate how the choice of regularization term influences the solution.