Probabilistic machine learning and phenomenological knowledge. Developments for optimization under uncertainty in safety-critical systems
Metadata
Show metadataAppears in the following Collection
- Matematisk institutt [3781]
Abstract
Artificial Intelligence (AI) and data-driven decisions based on Machine Learning (ML) are making an impact on an increasing number of industries. As these autonomous and self-learning systems become more and more responsible for decisions that may ultimately affect the safety of people, assets, or the environment, ensuring the safe use of AI will be crucial. This thesis aims to provide some of the tools needed to make data-driven modeling suitable for use in safety-critical systems, like a ship, offshore structure, or a spacecraft. This is challenging when we are faced with complex physical phenomena, in environments with a high degree of uncertainty, and where the consequence of an erroneous decision can be catastrophic. To succeed, the knowledge we possess about these phenomena must be exploited optimally. We consider various ways in which knowledge about the underlying physical system can be incorporated into probabilistic models. This includes how to make use of expensive computer simulations most efficiently, and how physics-based knowledge can be used as constraints to obtain “physically obedient machine learning models”. With this approach, we develop algorithms that can be used to search for optimal decisions in uncertain and safety-critical environments.List of papers
Paper I C. Agrell (2019). Gaussian Processes with Linear Operator Inequality Constraints. Journal of Machine Learning Research. Vol. 20, no. 135, pp. 1–36. The paper is included in the thesis in DUO. |
Paper II O. Gramstad, C. Agrell, E. Bitner-Gregersen, B. Guo, E. Ruth and E. Vanem (2020). Sequential sampling method using Gaussian process regression for estimating extreme structural response. Marine Structures. Vol. 72, 102780. The paper is included in the thesis in DUO, and also available at: https://doi.org/10.1016/j.marstruc.2020.102780 |
Paper III C. Agrell and K. R. Dahl (2021). Sequential Bayesian optimal experimental design for structural reliability analysis. Statistics and Computing. Vol. 31, no. 27. The paper is included in the thesis in DUO, and also available at: https://doi.org/10.1007/s11222-021-10000-2 |
Paper IV C. Agrell, K. R. Dahl and A. Hafver (2021). Optimal sequential decision making with probabilistic digital twins. Submitted for publication. arXiv: 2103.07405. To be published. The paper is removed from the thesis in DUO awaiting publishing. |
Paper V C. Agrell, S. Eldevik, O. Gramstad and A. Hafver (2021). Risk-based functional black-box optimization – Contribution to the NASA Langley UQ challenge on optimization under uncertainty. Mechanical Systems and Signal Processing. Vol. 164, 108266. The paper is included in the thesis in DUO, and also available at: https://doi.org/10.1016/j.ymssp.2021.108266 |
Paper VI A. Hafver, C. Agrell and E. Vanem (2021). Environmental contours as Voronoi cells. Published in Extremes Vol. 25, pp. 451–486 (2022). An author version is included in the thesis. The published version is available at: https://doi.org/10.1007/s10687-022-00437-7 |
Paper VII/Appendix C. Agrell, S. Eldevik, A. Hafver, F. B. Pedersen, E. Stensrud and A. Huseby (2018). Pitfalls of machine learning for tail events in high risk environments. Safety and Reliability — Safe Societies in a Changing World: Proceedings of ESREL 2018. pp. 3043–3051, CRC press. The paper is included in the thesis in DUO, and also available at: https://doi.org/10.1201/9781351174664-381 |