In this repo, some common used models are compared on a Navier Stokes Dataset.
The models, with their main characteristics and performances, are summed up here.
| Model | rel u train error | rel u test error super |
|---|---|---|
| POD+Tree | 1.8e-15 | 0.08 |
| POD DeepOnet (SpacetimeCoupled) | 0.73 | 0.52 |
| DeepOnet (SpacetimeCoupled) | 0.31 | 0.39 |
| FNO | 0.09 | 0.08 |
| AVNO | 0.20 | 0.16 |
| MIONet | 0.33 | 0.44 |
| DoublePOD-MIONet | 0.08 | 0.06 |
| DoublePOD-Tree | 0.05 | 0.16 |