A hierarchical greedy strategy for adaptive model-order reduction

Projection-based model-order reduction is a powerful methodology for solving parameter-dependent linear systems of equations. Adaptive multi-point methods commonly employ a greedy strategy for expansion point placement: The location where some error measure is maximum is selected. This requires evaluating an error indicator on a dense sampling of the parameter domain at each iteration of the model generation phase. To reduce runtimes, a hierarchical refinement strategy that reuses information from previous steps is proposed.