Inverse Reduced-Order Modeling
Résumé
We propose a general probabilistic formulation of reduced-order modeling in the case the system state is hidden and characterized by some uncertainty. The objective is to integrate noisy and incomplete observations in the process of building a reduced-order model. We call this problematic inverse reduced-order modeling. This problematic arises in many scientific domains where there exists a need of accurate low-order descriptions of highly-complex phenomena, which can not be directly and/or deterministically observed. Among others, it concerns geophysical studies dealing with image data, which are important for the characterization of global warming or the prediction of natural disasters.
Origine : Fichiers produits par l'(les) auteur(s)