New method trains autoregressive neural operators on solver-defined physical constraint manifolds
An arXiv paper proposes learning autoregressive neural operators that approximate time-dependent PDE solvers by composing a learned transition while keeping states on solver-defined physical constraint manifolds.
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-11-17 05:00 · arXiv stat.ML
Learning Autoregressive Neural Operators on Solver-Defined Physical Constraint Manifolds