Adaptive designs for multi-output polynomial chaos expansions and sensitivity analysis
METADATA ONLY
Loading...
Author / Producer
Date
2024-02
Publication Type
Other Conference Item
ETH Bibliography
yes
Citations
Altmetric
METADATA ONLY
Data
Rights / License
Abstract
Adaptive design of experiments has been demonstrated very effective in reducing the computational costs of complex uncertainty quantification tasks, such as reliability and sensitivity analysis. While traditionally associated with local- and kernel- based surrogate models like Gaussian process modelling and support vector machines, recent research has demonstrated that adaptive design of experiments can also benefit regression- based surrogate models, such as polynomial chaos expansions (PCEs). Nevertheless, one core limitation of most adaptive design strategies is that, with few exceptions, they are designed for scalar-output models only. When dealing with multiple output models, optimality conditions become much more difficult to define, and the literature on the subject is still sparse. In this contribution, we extend a recently proposed semi-supervised sequential design approach for sparse PCE, to the case of vector-output computational models. Thanks to the well-known synergy between PCE and Sobol' indices- based sensitivity analysis, this design of experiments strategy is also well suited for sensitivity analysis applications. We demonstrate the performance of this sequential design strategy on a number of well-known benchmarks from the surrogate modelling and sensitivity analysis literature.
Permanent link
Publication status
published
External links
Editor
Book title
SIAM Conference on Uncertainty Quantification (UQ 2024). Searchable Abstract Document
Journal / series
Volume
Pages / Article No.
122 - 123
Publisher
SIAM
Event
SIAM Conference on Uncertainty Quantification (UQ 2024)
Edition / version
Methods
Software
Geographic location
Date collected
Date created
Subject
Uncertainty quantification; Polynomial chaos expansions; Adaptive experimental designs; Global sensitivity analysis
Organisational unit
03962 - Sudret, Bruno / Sudret, Bruno
Notes
Conference lecture held on February 28, 2024.