The fast and the fewer

We will discuss the application of dimensionality reduction techniques to speed up computations and perform model forecast combination for Bayesian Uncertainty Quantification within nuclear physics. We use the same overarching framwork to both identify reduced coordinates for building single model emulators through model order reduction, and to combine groups of models within the Bayesian Model Mixing approach. Both directions are quite important for fostering the experiment/observaiton/theory cycle with well quantified uncertainties.