Get Started
Home
Topics
Search
Library
Research questionHow can periodic PINNs reduce the cost of high-order spatial derivatives without sacrificing solution accuracy?PDE residuals can require multiple or high-order spatial derivatives, making their computation costly in both training time and memory. Periodic PINNs must also account for how the differentiation procedure interacts with the available spatial representation.
AI
Evaluation & Benchmarks
Machine Learning
Neural and Evolutionary Computing
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.A Computational Comparison of Fourier Spectral Differentiation and Spatial Automatic Differentiation in Periodic Physics-Informed Neural NetworksThe evidence compares spatial automatic differentiation with Fourier spectral differentiation on Allen–Cahn, Korteweg–de Vries, and Kuramoto–Sivashinsky benchmarks. Neural representations, temporal differentiation, optimization, sampling, and training schedules were paired and held fixed; the Fourier approach requires a uniform structured spatial grid. Results cover one-dimensional periodic settings and report training speed, peak GPU memory, and relative L2 error.research paper · Sep 2, 2026
Related questions
How can PINN transfer learning recover physical parameters when source and target PDEs differ?How can residual networks approximate high-dimensional semilinear heat-equation solutions without exponential parameter growth?How can parametric PDE solution operators extrapolate to unseen regimes without failing silently?How can DeepONet-style neural operators represent sharp moving discontinuities in shock-dominated, low-viscosity PDEs?