EasyVVUQ: toolkit for Verification, Validation and Uncertainty Quantification

Enables automated VVUQ that can be deployed on emerging exascale platforms for many domains of interest.

EasyVVUQ is an open-source Python library mainly designed to facilitate forward parametric uncertainty quantification (UQ) for a wide variety of simulations. Most computer models are uncertain, e.g. they depend upon a number of tunable input parameters, which are almost never known exactly. Forward UQ is the process of propagating this input uncertainty through the computer models in order to examine the corresponding uncertainty in the predictions of that model, which is important to assess the confidence we can have in the simulations. This invariably requires ensembles, i.e. running the same model multiple times at different (parameter) settings.
The goal of EasyVVUQ is to make it as easy as possible to implement various advanced forward UQ techniques for existing application codes or workflows.

EasyVVUQ was originally developed in the EU VECMA project (www.vecma.eu), which was an exascale computing project. As such, another key consideration is the ability to support high-performance computing (HPC) codes. This is achieved by integrating EasyVVUQ with other VECMA libraries (FabSim3 and QCG PilotJob), which allows to execute large EasyVVUQ ensembles on remote supercomputers.

Currently EasyVVUQ is maintained at CWI, in collaboration with different partners from the UK-based SEAVEA project (https://www.seavea-project.org/).