Chao Dang, Insitute for Risk and Reliability, Leibniz Universität Hannover, Germany
Forward Uncertainty Quantification with Special Emphasis on a Bayesian Active Learning Perspective (2023)
Supervisor: Prof. Dr.-Ing. Michael Beer
Keywords: Uncertainty propagation; Structural reliability analysis; Bayesian active learning; Bayesian inference; Bayesian quadrature; Bayesian optimization; Active learning; Gaussian process; Numerical uncertainty; Line sampling
Forward uncertainty quantification (UQ) plays an important role in understanding the effects of the input uncertainties of a computational model on the outputs, which in turn enables more informed decision-making, risk management, and improved predictions or results. This thesis focuses on developing innovative approaches for efficient and accurate forward UQ, especially from a machine learning paradigm called Bayesian active learning. The underlying idea is to formulate a forward UQ task as a Bayesian inference problem, and design essential components (e.g., learning function and stopping criterion) for active learning using the quantified uncertainty of the quantity of interest.
More information on Chao's work: