| Experiment: gin+selfies on bindingdb |
|
|
| FOLD 1/5 |
| GNN Train Loss: 0.9757 |
| RMSE: 1.5584, MAE: 1.2829, R2: 0.0512 |
|
|
| FOLD 2/5 |
| GNN Train Loss: 0.9506 |
| RMSE: 1.4919, MAE: 1.2230, R2: 0.0243 |
|
|
| FOLD 3/5 |
| GNN Train Loss: 0.9521 |
| RMSE: 1.4578, MAE: 1.2127, R2: 0.0321 |
|
|
| FOLD 4/5 |
| GNN Train Loss: 0.9585 |
| RMSE: 1.5511, MAE: 1.2722, R2: 0.0301 |
|
|
| FOLD 5/5 |
| GNN Train Loss: 0.9451 |
| RMSE: 1.4885, MAE: 1.2289, R2: 0.0266 |
|
|
| ====== K-FOLD SUMMARY ====== |
| RMSE: 1.5105 [1.4769, 1.5429] |
| MAE: 1.2441 [1.2201, 1.2699] |
| R2: 0.0331 [0.0263, 0.0428] |
|
|
| ★ Using fold 3 model for test inference |
| RMSE: 1.7509, MAE: 1.3484, R2: 0.0653 |
|
|
| ====== HELD-OUT TEST METRICS ====== |
| {'rmse': 1.7508875094999254, 'mae': 1.348360538482666, 'r2': 0.06534677743911743} |
| Test RMSE: 1.7480 (95 % CI: 1.5285–2.0136) |
|
|
| Test MAE : 1.3502 (95 % CI: 1.2410–1.4656) |
|
|
| Test R² : 0.0639 (95 % CI: 0.0251–0.1043) |
|
|
|
|
