Regression on Synthetic two-feature data Square, n=1000
0.9965Mean R^2LightGBM, rel-s features
Evaluation results
| Method | Links | ||
|---|---|---|---|
| LightGBM, rel-s featuresSample size (n)=10002025.12 | 0.9965 | 0.0005 | |
| NW + relationsNorm/Embedding Type=L2 norm, Sample size (n)=10002025.12 | 0.9874 | 0.0013 | |
| NW + relationsNorm/Embedding Type=learnable norm, Sample size (n)=10002025.12 | 0.9679 | 0.0028 | |
| NW + relationsNorm/Embedding Type=MLP embeddings, Sample size (n)=10002025.12 | 0.9679 | 0.0022 | |
| TabRelSample size (n)=10002025.12 | 0.9279 | 0.0311 | |
| NW, no rel-s infoNorm/Embedding Type=learnable norm, Sample size (n)=10002025.12 | 0.6623 | 0.0289 | |
| NW, no rel-s infoNorm/Embedding Type=L2 norm, Sample size (n)=10002025.12 | 0.6594 | 0.0297 | |
| LightGBM, no rel-s infoSample size (n)=10002025.12 | 0.6056 | 0.0392 | |
| NW, rel-s featuresSample size (n)=10002025.12 | -0.902 | 0.1244 |