May 8-10

Abstract Details

files Add files

Abstracts

Author: Scott Kruger
Requested Type: Poster
Submitted: 2023-05-02 13:07:23

Co-authors: S. Kruger [Tech-X], J. Leddy, E. Howell [Tech-X], S. Madireddy [ANL], L.L. Lao, C. Akcay, T.A. Bechtel, Y.Q. Liu, J. McClenaghan, D. Orozco, D. Schissel, S.P. Smith, X. Sun [General Atomics], S. Williams, O. Antepara [LBNL], A. Pankin [PPPL]

Contact Info:
Tech-X
5621 Arapahoe Ave
Boulder, Colorado   80303-1378
United States

Abstract Text:
Bayesian statistics is a powerful tool for making inferences about the world based on data. It is particularly well-suited for scientific applications, as it naturally treats everything as a probability. This perspective allows for uncertainty quantification, incorporation of prior knowledge, and robustness to ill-posed problems. Here, we talk about how this approach relates to "equilibrium reconstruction", such as that performed by EFIT-AI, and "interpretative transport codes", such as that performed by TRANSP. We contrast these approaches to the integrated data analysis approach being investigated by European physicists. We discuss the strengths and weaknesses, and how machine learning addresses many of the limitations of the Bayesian approach.

Comments: