Error-Conditioned Neural Solvers

DOI10.48550/arXiv.2606.27354MaRDI QIDQ7270117arXivFDO

Authors Haina Jiang, Seungryong Kim, Liam Wang, Peng-Chen Chen, Min Seop Kwak, Brian Bell, Jeong Joon Park

Publication date 25 June 2026

Published in arXiv

Tldr This paper introduces error-conditioned Neural Solvers (ENS), a novel approach to leveraging neural networks for solving PDEs that conditions on the spatial structure of its own errors rather than optimizing an objective function based on residuals, achieving higher accuracy and generalization across multiple PDE families.

Full work available at URL https://doi.org/10.48550/arXiv.2606.27354

arXiv classification

cs.AIcs.CVcs.LGmath.NA

This page was built for publication: Error-Conditioned Neural Solvers

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7270117)