Farid Mohammadi, former doctoral researcher at the Department of Hydromechanics and Modelling of Hydrosystems (SFB 1313 research project D03), and member within SFB 1313 Integrated Research Training Group IRTG-IMPM, will defend his dissertation:
Title: "A Surrogate-Assisted Bayesian Framework for Uncertainty-Aware Validation Benchmarks"
Date: 14 February 2023
Time: 11 am CET
Venue: Department of Hydromechanics and Modelling of Hydrosystems, Pfaffenwaldring 61, 70569 Stuttgart, room: MML
Abstract
In many modeling tasks, our limited knowledge about the interaction of the governing processes of investigated systems leads to several modeling approaches. These modeling concepts include different levels of detail and different assumptions and can be compared as competing hypotheses in validation benchmarks. In my talk, I will introduce a two-stage Bayesian multi-model framework to help modelers perform uncertainty-aware model validation benchmarks.
To keep the computational costs reasonable, a surrogate model based on polynomial chaos expansion has been used to accelerate the analyses for computationally demanding models. Moreover, I will discuss how surrogate models could benefit from sparse representation and sequential learning to achieve more accurate predictions with as few simulations as the computational budget allows.
The surrogate-assisted uncertainty-aware validation framework was applied to a range of benchmark studies to investigate flow and transport in porous media. It can, however, be applied to other disciplines where models are used to make predictions.