« Context-Aware Model Reduction »
Event details
| Date | 30.10.2018 |
| Hour | 16:15 › 17:15 |
| Speaker | Prof. Benjamin Peherstorfer |
| Location | |
| Category | Conferences - Seminars |
Abstract:
Traditional model reduction speeds up computations by replacing expensive high-fidelity models with low-cost reduced models. Thus, the goal of traditional model reduction is to construct reduced models that approximate well the high-fidelity models; however, reduced models are increasingly used together with high-fidelity models, which means that the primary purpose of reduced models becomes supporting computations with the high-fidelity models rather than approximating and replacing them. In this presentation, we propose context-aware model reduction strategies that explicitly construct reduced models for being used together with other models. In the first part, we present an approach to dynamically couple reduced models with the high-fidelity model, where the reduced models are adapted in a context-aware sense with sparse data from the high-fidelity model. Our numerical examples demonstrate that the dynamic coupling is particularly beneficial in case of transport-dominated problems, where our context-aware approach achieves significant speedups, whereas traditional reduced models are even more costly to evaluate than the high-fidelity models. In the second part of the presentation, we introduce the adaptive multifidelity Monte Carlo (AMFMC) method that constructs reduced models that support the multifidelity estimation of statistics of high-fidelity model outputs. Our analysis shows that our context-aware reduced models optimally reduce the runtime of the multifidelity estimation, even though the context-aware reduced models are less accurate in the sense of traditional model reduction.
Practical information
- General public
- Free
Organizer
- MCSS
Contact
- Delphine Vieira