Reproduce “Operator Learning Using Weak Supervision from Walk-on-Spheres” 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
Hrishikesh Viswanath, Hong Chul Nam, Xi Deng, Julius Berner, Anima Anandkumar, Aniket Bera
WoS-NO trains reusable neural operators for PDE families by matching predictions to noisy solution estimates from random walks, avoiding precomputed solution datasets and PDE-derivative losses. On linear Poisson problems, it reports mean MSE of 8.2 × 10⁻⁴ after 20,000 steps, versus PINO’s 2.5 × 10⁻³. For varying-coefficient problems, accuracy is similar to PINO’s; the evaluation is limited in scope.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.