PapersSelected73°

Euclidean Fourier Neural Operators

Why we picked it

EFNOs solve a fundamental problem in scientific computing by enabling models to generalize across different domains without retraining, which could accelerate materials discovery across ASEAN research institutions.

Euclidean Fourier Neural Operators (EFNOs) address a key limitation of traditional Fourier neural operators (FNOs) by being domain-independent.

Body

FNOs struggle with transfer across different domains due to their dependence on periodic domains and integer Fourier mode numbers.

EFNOs parameterize the spectral kernel as a continuous function of physical wavevectors, enabling consistent operation across periodic domains of varying shapes and sizes.

The researchers evaluated EFNOs on a heat equation and a materials science task involving exchange-correlation potentials across different crystal structures.

Original · arXiv cs.LG

Fourier neural operators (FNOs) provide an efficient framework for learning mappings between function spaces as they are, by construction, independent of the grid resolution at which they are trained and evaluated. However, FNOs are not independent of the periodic domain they are applied to: their discrete spectral weights are indexed by integer Fourier mode numbers, which correspond to physical wavevectors. When applied to a different domain, the same trained weights act at different wavevectors, and the FNO silently represents a different operator. This makes FNOs unsuitable for tasks where transfer across domains is crucial. We propose Euclidean Fourier neural operators~(EFNOs) as a domain-independent alternative to FNOs. By parameterizing the spectral kernel as a continuous function of the physical wavevector, the EFNO can learn operators that act consistently across periodic domains of varying shape and size. We evaluate the EFNO on a simple heat equation and on a practically relevant materials science task of learning exchange-correlation potentials across different crystal structures, and demonstrate that the EFNO is able to generalize to unseen grid sizes and domains.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "Papers" desk. Key entities and technical topics under continuous tracking include EFNOs, FNOs, Fourier Neural Operators, materials science.

Source attribution: curated from arXiv cs.LG (@Nathanael Bosch, Niklas Frederik Schmitz, Michael F. Herbst).