Adaptive anisotropic composite quadratures for residual minimisation in neural PDE approximations

Santiago Badia †, Kishore Nori † · 2026

Read Adaptive anisotropic composite quadratures for residual minimisation in neural PDE approximations by Santiago Badia †, Kishore Nori † (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-00308.txt.

Description

Santiago Badia and Kishore Nori study a hidden ingredient in neural methods for partial differential equations: the numerical quadrature behind the training loss. These methods train a network by minimizing a residual, and the residual must be estimated at sample points. Errors at that stage can quietly cap the final accuracy, whatever the network's power.

The authors first split the total error into approximation, quadrature, and optimization parts, with an estimate of how quadrature error infects the result. Guided by that analysis, they propose an adaptive composite quadrature that refines points where the error indicator runs high, rebuilding the grid only when needed. On benchmark problems, the approach narrowed the gap between training and reference losses, used points more efficiently, and beat fixed quadrature strategies on accuracy.

Catalog record

CREATORSantiago Badia †, Kishore Nori †
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE104 KB
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.00308

Browse the full library · About the Tower of Babel