Neural Super-Resolution for Mesh-Based Simulations on Dataset 3
0.0243RMSEFully supervised
Evaluation Results
| Method | Links | |
|---|---|---|
| Fully supervisedNh (number of HR data)=200, N (number of low-resolution data)=200, Inductive biases=without inductive biases, Supervision type=full supervision, Base architecture=MGN2026.05 | 0.0243 | |
| Fully supervisedNh, N=200, 200, MPNN Architecture=MGN2026.05 | 0.0243 | |
| SuperMeshNetNh (number of HR data)=20, N (number of low-resolution data)=200, Inductive biases=with inductive biases, Supervision type=partial supervision, Base architecture=MGN2026.05 | 0.0245 | |
| SuperMeshNetNh, N=20, 200, MPNN Architecture=MGN2026.05 | 0.0245 | |
| SuperMeshNetNh, N=20, 200, MPNN Architecture=GIN2026.05 | 0.0258 | |
| SuperMeshNet-ONh, N=20, 200, MPNN Architecture=MGN2026.05 | 0.0281 | |
| SuperMeshNetNh, N=20, 200, MPNN Architecture=GTR2026.05 | 0.0294 | |
| SuperMeshNetNh, N=20, 200, MPNN Architecture=GCN2026.05 | 0.0297 | |
| SuperMeshNetNh, N=20, 200, MPNN Architecture=SAGE2026.05 | 0.0297 | |
| SuperMeshNetNh, N=20, 200, MPNN Architecture=GAT2026.05 | 0.031 | |
| SuperMeshNet-ONh, N=20, 200, MPNN Architecture=GIN2026.05 | 0.0316 | |
| Fully supervisedNh, N=200, 200, MPNN Architecture=GIN2026.05 | 0.0317 | |
| Fully supervisedNh, N=200, 200, MPNN Architecture=GTR2026.05 | 0.0329 | |
| Fully supervisedNh, N=200, 200, MPNN Architecture=SAGE2026.05 | 0.034 | |
| SuperMeshNet-ONh, N=20, 200, MPNN Architecture=GTR2026.05 | 0.0363 | |
| SuperMeshNet-ONh, N=20, 200, MPNN Architecture=SAGE2026.05 | 0.0366 | |
| Fully supervisedNh, N=200, 200, MPNN Architecture=GCN2026.05 | 0.037 | |
| Fully supervisedNh, N=200, 200, MPNN Architecture=GAT2026.05 | 0.0374 | |
| SuperMeshNet-ONh, N=20, 200, MPNN Architecture=GAT2026.05 | 0.0375 | |
| SuperMeshNet-ONh, N=20, 200, MPNN Architecture=GCN2026.05 | 0.038 | |
| Fully supervisedNh, N=20, 20, MPNN Architecture=GTR2026.05 | 0.0513 | |
| Fully supervisedNh, N=20, 20, MPNN Architecture=MGN2026.05 | 0.0523 | |
| Fully supervisedNh, N=20, 20, MPNN Architecture=GIN2026.05 | 0.0569 | |
| Fully supervisedNh, N=20, 20, MPNN Architecture=GCN2026.05 | 0.0587 | |
| Fully supervisedNh, N=20, 20, MPNN Architecture=SAGE2026.05 | 0.0611 | |
| Fully supervisedNh, N=20, 20, MPNN Architecture=GAT2026.05 | 0.0616 | |
| MAgNetNh (number of HR data)=0, N (number of low-resolution data)=200, Supervision type=no HR supervision (zero-shot)2026.05 | 0.0754 |