Deep collocation method: A framework for solving PDEs using neural networks with error control

Mingxing Weng, Zhiping Mao, Jie Shen · 2025

Read Deep collocation method: A framework for solving PDEs using neural networks with error control by Mingxing Weng, Zhiping Mao, Jie Shen (2025) 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-2502-17203.txt.

Description

Mingxing Weng, Zhiping Mao, and Jie Shen present a neural method for solving physics equations with built-in error checks, in a paper from twenty twenty-five. Networks can approximate hard functions, but one-shot training often lacks accuracy and wastes effort. Their method grows the solution step by step instead.

Small single-layer networks supply building-block functions, steered by how far the current answer misses the equation. The answer is then fit inside the span of those blocks by least squares. Each round's gain doubles as an error gauge that says when to stop. Smart choices of sample points and starting values add robustness. Tests on hard equations show high accuracy, and the authors supply error bounds to match.

Catalog record

CREATORMingxing Weng, Zhiping Mao, Jie Shen
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE67 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-17203

Browse the full library · About the Tower of Babel