2 papers
physics.plasm-ph2025
Efficient dataset construction using active learning and uncertainty-aware neural networks for plasma turbulent transport surrogate models
Aaron Ho, Lorenzo Zanisi, Bram de Leeuw +3
This work demonstrates a proof-of-principle for using uncertainty-aware architectures, in combination with active learning techniques and an in-the-loop physics simulation code as…
physics.plasm-ph2025
Duqtools: Dynamic uncertainty quantification for Tokamak reactor simulations modelling
Victor Azizi, Stef Smeets, Florian Koechl +3
Large scale validation and uncertainty quantification are essential in the experimental design, control, and operations of fusion reactors. Reduced models and increasing computatio…