The Python Toolkit for Uncertainty Quantification (PyTUQ) is a lightweight Python library for a range of uncertainty quantification tasks and workflows. Features include conventional tools such as polynomial chaos machinery with mixed bases, global sensitivity analysis, quadrature point generation, linear regression, Bayesian inference with various flavors of Markov chain Monte Carlo. PyTUQ also includes advanced methods such as Bayesian compressed sensing, sampling-based Rosenblatt transformation and embedded model error calibration.
In this work, we will present the structure of the library, provide several educational examples, and demonstrate PyTUQ’s application to various scientific problems, ranging from Earth system modeling to chemistry and materials science.