arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2408.04846math.NAcs.AIcs.LGcs.MScs.NA
收藏

UGrid:一种用于线性 PDE 的高效且严谨的神经多重网格求解器

UGrid: An Efficient-And-Rigorous Neural Multigrid Solver for Linear PDEs

Stony Brook University (SUNY)(纽约州立大学石溪分校);Institute of Software, Chinese Academy of Sciences(中国科学院软件研究所);University of Chinese Academy of Sciences(中国科学院大学)

Xi Han, Fei Hou, Hong Qin

AI总结:

本文提出融合 U-Net 与多重网格的线性 PDE 神经求解器 UGrid,兼具收敛与正确性证明、高精度及跨几何和 PDE 形式的泛化能力,并以残差损失实现稳定无监督训练。

摘要:

偏微分方程(PDE)的数值求解器对科学与工程具有根本性意义。迄今为止,历史上对传统技术的依赖限制了与大数据知识的可能融合,并且在某些 PDE 形式下表现出次优效率;而数据驱动的神经方法通常缺乏关于收敛性和正确性的数学保证。本文阐明了一种用于线性 PDE 的数学严谨的神经求解器。所提出的 UGrid 求解器建立在 U-Net 与 MultiGrid(多重网格)原则性融合的基础上,给出了关于收敛性和正确性的数学严谨证明,并展现出高数值精度,以及对各种输入几何/数值和多种 PDE 形式的强泛化能力。此外,我们设计了一种新的残差损失指标,它支持无监督训练,并相较传统损失提供更高稳定性和更大的解空间。

英文摘要:

Numerical solvers of Partial Differential Equations (PDEs) are of fundamental significance to science and engineering. To date, the historical reliance on legacy techniques has circumscribed possible integration of big data knowledge and exhibits sub-optimal efficiency for certain PDE formulations, while data-driven neural methods typically lack mathematical guarantee of convergence and correctness. This paper articulates a mathematically rigorous neural solver for linear PDEs. The proposed UGrid solver, built upon the principled integration of U-Net and MultiGrid, manifests a mathematically rigorous proof of both convergence and correctness, and showcases high numerical accuracy, as well as strong generalization power to various input geometry/values and multiple PDE formulations. In addition, we devise a new residual loss metric, which enables unsupervised training and affords more stability and a larger solution space over the legacy losses.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑