Reproduce “Adaptive Domain Decomposition for Finite Basis Physics-Informed Neural Networks” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Sarah Helfert, Wanxin Chen, Amanda A. Howard, Panos Stinis, Alexander Heinlein

ADD-FBPINN adaptively learns local subdomains for physics-informed neural PDE solvers. It trains a global PINN, then uses its predictions to guide learned window functions before training an FBPINN. In the tested anisotropic two-dimensional Helmholtz case, its relative L2 error was 4.73×10⁻³, versus 9.90×10⁻¹ for a PINN and 6.02×10⁻² for a fixed uniform FBPINN with nine subdomains; this demonstrates results on tested problems, not general performance.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.