CAFE+FNO: Fourier Kernel Generation via Multiplicative Feature Composition

DOI10.48550/arXiv.2610.10105MaRDI QIDQ7394871arXivFDO

Authors Minwoo Shin, Hyungjoon Juen

Publication date 7 October 2026

Published in arXiv

Tldr We propose a new Fourier kernel generation method that combines the strengths of both spatially varying kernel learning and frequency interactions through explicit feature composition. CAFE+FNO learns to construct complex channel-mixing matrices for each normalized spectral coordinate, enabling high-quality learning of PDE solution operators with fewer parameters than existing methods.

Full work available at URL https://doi.org/10.48550/arXiv.2610.10105

arXiv classification

cs.LGmath.NA

Cites work

• Choose a Transformer: Fourier or Galerkin
• Fourier Neural Operator for Parametric Partial Differential Equations
• Fourier Neural Operator with Learned Deformations for PDEs on General Geometries
• Frequency Principle: Fourier Analysis Sheds Light on Deep Neural Networks
• GNOT: A General Neural Operator Transformer for Operator Learning
• Model reduction and neural networks for parametric PDEs
• Multipole Graph Neural Operator for Parametric Partial Differential Equations
• Neural Operator: Graph Kernel Network for Partial Differential Equations
• Neural Operator: Learning Maps Between Function Spaces
• Tensor Decompositions and Applications
• Tensor-train decomposition
• The Random Feature Model for Input-Output Maps between Banach Spaces

This page was built for publication: CAFE+FNO: Fourier Kernel Generation via Multiplicative Feature Composition

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7394871)