Table 2 Results of training on the secretory dataset using different pooling mechanisms, validated on the secretory test set
From: AMPDeep: hemolytic activity prediction of antimicrobial peptides using transfer learning
Pooler | Hidden layer size | Accuracy | Recall | Precision | ROC-AUC | MCC |
|---|---|---|---|---|---|---|
BERT Pooler | None | 0.9431 | 0.9624 | 0.9275 | 0.9795 | 0.8868 |
Mean Pooling | 32 | 0.9458 | 0.9731 | 0.9235 | 0.9827 | 0.8929 |
First Token Pooling | 128 | 0.9566 | 0.957 | 0.957 | 0.9779 | 0.9133 |