Reproduce “Learnable Splines for Unstructured Neural Operators” 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
Vignesh Gopakumar, Zhuoyuan Wang, Ander Gray, Valentin Duruisseaux, Josh Williams, William Michael Edward Ellis, Nuno Nobre, Matthew Field and 7 more
L-SplineFNO adapts Fourier neural operators to irregular engineering meshes by scattering mesh features onto a regular grid, processing them there, and reconstructing fields with a differentiable spline decoder. On Heat Sink, it reports 0.027 ± 0.001% relative L2 error and 00:42 training, versus 0.035 ± 0.001% and 01:22 for analytic SplineFNO. Accuracy varies by task; bracket stresses are an exception.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.