A Training Set Subsampling Strategy for the Reduced Basis Method

Abstract We present a subsampling strategy for the offline stage of Reduced Basis Method. The approach is aimed at bringing down considerable costs associated with using finely-sampled training set. proposed algorithm exploits potential pivoted QR decomposition and discrete empirical interpolation method to identify important parameter samples. It consists two stages. In first stage, we construct low-fidelity approximation solution manifold over fine Then, available snapshots output variable, apply or set sparse sampling locations in domain. These points reveal structure parametric dependence variable. second proceeds subsampled containing by far smaller number parameters than initial Different strategies inspired from recent variants are also considered. Tests on benchmark examples justify new show its substantially speed up Method, while generating reliable reduced-order models.