Reproduce “Predictive Uncertainty for Neural CAE Surrogates” 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
Kaustubh Tangsali, Mohammad Amin Nabian, Kelvin Lee, Carmelo Gonzales, Sanjay Choudhry
The authors compare three uncertainty methods for neural CAE surrogates across aerodynamic and crash simulations. On DrivAerStar, all three rank local errors usefully, and validation-fitted scaling gives near-nominal intervals on disjoint Fastback cases. Under an Estateback shift, GP-total pressure error grows 1.235 times while predicted standard deviation grows 1.039 times; uncertainty therefore does not track shifted error proportionally.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.