Neural Preconditioning Operator for Efficient PDE Solves

Zhihao Li, Di Xiao, Zhilu Lai, Wei Wang · 2025

Read Neural Preconditioning Operator for Efficient PDE Solves by Zhihao Li, Di Xiao, Zhilu Lai, Wei Wang (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-2502-01337.txt.

Description

Zhihao Li and co-authors present a neural preconditioning operator that speeds solvers for partial differential equations. Large sparse systems from such equations dominate simulation cost. Classic preconditioners need heavy tuning and transfer poorly across meshes.

Their operator learns from condition and residual losses during training. It plugs into standard Krylov solvers as a learned helper. Multigrid ideas blend with a transformer design inside it. Tests span heat, diffusion, and elasticity problems. Both even and rough meshes were included in the trials. Iteration counts and runtimes fell against classic methods. Convergence held on grids far larger than the training ones. The work pushes data-driven help for core solvers.

Catalog record

CREATORZhihao Li, Di Xiao, Zhilu Lai, Wei Wang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-01337

Browse the full library · About the Tower of Babel