Message Passing Neural PDE Solvers
About
The numerical solution of partial differential equations (PDEs) is difficult, having led to a century of research so far. Recently, there have been pushes to build neural--numerical hybrid solvers, which piggy-backs the modern trend towards fully end-to-end learned systems. Most works so far can only generalize over a subset of properties to which a generic solver would be faced, including: resolution, topology, geometry, boundary conditions, domain discretization regularity, dimensionality, etc. In this work, we build a solver, satisfying these properties, where all the components are based on neural message passing, replacing all heuristically designed components in the computation graph with backprop-optimized neural function approximators. We show that neural message passing solvers representationally contain some classical methods, such as finite differences, finite volumes, and WENO schemes. In order to encourage stability in training autoregressive models, we put forward a method that is based on the principle of zero-stability, posing stability as a domain adaptation problem. We validate our method on various fluid-like flow problems, demonstrating fast, stable, and accurate performance across different domain topologies, equation parameters, discretizations, etc., in 1D and 2D.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Zero-shot super-resolution | E1 (test) | MAE0.1541 | 48 | |
| Learning PDE Solution Operators | 2D Shallow Water | Mean L2 Relative Error0.5 | 20 | |
| Surrogate modeling of 1D nonlinear PDEs | 1D family of nonlinear PDEs Scenario E1 1.0 (test) | Accumulated Error (MSE)0.51 | 15 | |
| Surrogate modeling of 1D nonlinear PDEs | 1D family of nonlinear PDEs Scenario E3 1.0 (test) | Accumulated Error (MSE)3.7 | 15 | |
| Surrogate modeling of 1D nonlinear PDEs | 1D family of nonlinear PDEs Scenario E2 1.0 (test) | Accumulated Error (MSE)1.45 | 15 | |
| Learning PDE Solution Operators | 1D Diffusion-Reaction | Mean L2 Rel Error39 | 12 | |
| Learning PDE Solution Operators | Allen-Cahn 1D | Mean L2 Relative Error0.33 | 12 | |
| Learning PDE Solution Operators | 1D Cahn-Hilliard | Mean L2 Relative Error0.59 | 8 | |
| Learning PDE Solution Operators | 1D Diffusion-Sorption | Mean L2 Relative Error2.86 | 8 | |
| Learning PDE Solution Operators | 3D Maxwell | Mean L2 Relative Error0.7445 | 8 |