Table 2 The prediction effect of label smoothing (LS) strategy on three tasks of Dataset1
From: MDDI-SCL: predicting multi-type drug-drug interactions via supervised contrastive learning
| Â | ACC | AUPR | AUC | F1 | Precision | Recall |
|---|---|---|---|---|---|---|
Task1 | ||||||
 With LS | 0.9378 | 0.9782 | 0.9983 | 0.8755 | 0.8804 | 0.8767 |
 Without LS | 0.9377 | 0.9776 | 0.9981 | 0.8840 | 0.8718 | 0.9023 |
Task2 | ||||||
 With LS | 0.6767 | 0.6947 | 0.9634 | 0.5304 | 0.6254 | 0.4814 |
 Without LS | 0.6659 | 0.6705 | 0.9470 | 0.5120 | 0.5275 | 0.5243 |
Task3 | ||||||
 With LS | 0.4589 | 0.3938 | 0.9053 | 0.1919 | 0.2585 | 0.1678 |
 Without LS | 0.4449 | 0.3636 | 0.8723 | 0.1971 | 0.2022 | 0.2063 |