Probabilistic Predictions With Fourier Neural Operators
MCML Authors
Philipp Scholl
Gitta Kutyniok
Prof. Dr.
Core PI
Abstract
Philipp Scholl
Gitta Kutyniok
Prof. Dr.
Core PI
Abstract
Neural networks have been successfully applied in modeling partial differential equations, especially in dynamical systems. Commonly used models, such as neural operators, are performing well at deterministic prediction tasks, but lack a quantification of the uncertainty inherent in many complex systems, for example weather forecasting. In this paper, we explore a new approach that combines Fourier neural operators with generative modeling based on strictly proper scoring rules in order to create well-calibrated probabilistic predictions of dynamical systems. We demonstrate improved predictive uncertainty for our approach, especially in settings with very high inherent uncertainty.
inproceedings BSK24
BDU @NeurIPS 2024
Workshop Bayesian Decision-making and Uncertainty: from probabilistic and spatiotemporal modeling to sequential experiment design at the 38th Conference on Neural Information Processing Systems. Vancouver, Canada, Dec 10-15, 2024.Authors
C. Bülte • P. Scholl • G. KutyniokLinks
URLResearch Area
BibTeXKey: BSK24