This workshop is aimed at the participants in the Trimester Program "Computational multifidelity, multilevel, and multiscale methods". It is not possible to apply only for this workshop.
Organizers:
- Niklas Baumgarten (Heidelberg University)
- Tung Le (Carl von Ossietzky Universität Oldenburg)
- Emil Løvbak (Karlsruhe Institute of Technology)
- Anastasia Istratuca (University of Edinburgh and Heriot-Watt University)
- Pieterjan Robbe (Sandia National Laboratories)
- Linus Seelinger (Karlsruhe Institute of Technology)
- Tommaso Vanzan (Politecnico di Torino)
- Steffen W. R. Werner (Virginia Polytechnic Institute and State University)
Description:
Many simulation problems contain phenomena at multiple scales of resolution in terms of both space and time. In practice, solving such problems is often challenging due to inherently stochastic model behavior or uncertain parameters, modeled probabilistically. As such, the demand for sophisticated techniques for stochastic simulation and uncertainty quantification (UQ) is growing across scientific and engineering disciplines. Hierarchical methods, which use multilevel and multifidelity approaches, are particularly promising, balancing accuracy and computational efficiency.
In this workshop we invite contributions from researchers working in the following thematic areas:
- Multilevel and Multifidelity methods: Multilevel and multifidelity strategies judiciously allocate computational resources between expensive high-fidelity models and less costly low-fidelity models. These strategies allow for the efficient computation of low variance/high accuracy results. Hence, efficient simulation and UQ are possible.
- Multifidelity Surrogate Models: Resource constraints often limit data availability for computationally expensive simulations. Multifidelity surrogate models, such as multifidelity Gaussian processes, chain models with increasing fidelity, balancing error and cost. These surrogates support adaptive sampling strategies, incrementally improving accuracy.
- Applications of Multifidelity Methods: Multifidelity techniques are applied across disciplines, from fusion energy to climate modeling. This focus area will feature domain experts demonstrating the use of these tools in real-world scenarios, showcasing their ability to provide accurate, reliable UQ. This includes software packages for multifidelity stochastic simulation and UQ.