Join
PDEBench fluid evolution
Neural surrogates that roll out PDE solutions with less error (placeholder)
| # | nRMSE ↓ | Step | What changed |
|---|---|---|---|
| 5 | 0.0684 | ↓ 0.0047 | A later commit on this route |
| 4 | 0.0731 | ↓ 0.0067 | Push-forward training on unrolled steps |
| 3 | 0.0798 | ↓ 0.0036 | A later commit on this route |
| 2 | 0.0834 | ↓ 0.0028 | Project each step onto mass and energy conservation |
| 1 | 0.0862 | ↓ 0.0048 | A multi-scale U-net with modern blocks |
| 0 | 0.091 | The baseline, FNO |
