Speaker
Description
Reduced order modeling methods characterized by significant reduction of computational time without loss of accuracy are extremely valuable for solution of large scale forward and inverse problems. The talk discusses a group of model order reduction methods related to matrix rational function (and matrix Pade) approximations of the spectral representations arising in inverse problems. The methods are based on data optimization and result in the reduced solution basis recovering the most relevant components of the solution. The approach is based on the minimization of the maximal approximation error and allows to significantly reduce the time and memory needed for solving the large scale problems. We show applications to inverse conductivity, scattering, and inverse Maxwell's problem.