Figure 3
From: Multi-task learning to leverage partially annotated data for PPI interface prediction
Comparison of the single-task model and multi-task models based on the AUC ROC scores of the PPI interface prediction on the validation and an independent test set. Performances are shown for the validation (dark blue) and test (blue) set for models trained on the PPI dataset, and the validation (dark red) and test (red) set for models trained on the PPI_extendedSFD dataset. All models are trained once on the training set. Similar performances are shown for the validation and test set. The multi-task models outperform the single-task model.
