MHNfs is a few-shot method trained on FS-Mol. The trained model is applied to the benchmark by using the training data as a support set without updating any model weigths.
No tags found.
| Test set | Target label | Metric | Score |
|---|---|---|---|
| test | CLASS_EGFR | f1 | 0.2191780821917808 |
| test | CLASS_EGFR | mcc | 0.11602387022306428 |
| test | CLASS_EGFR | pr_auc | 0.37336228322821635 |
| test | CLASS_EGFR | roc_auc | 0.8125 |
| test | CLASS_EGFR | accuracy | 0.20833333333333334 |
| test | CLASS_EGFR | cohen_kappa | 0.026565464895635715 |