Closed aoifecahill closed 4 years ago
Currently, we allow for all variants of the f1 metric in metrics.py, but we do not do the same for precision and recall.
This means that we cannot tune on precision/recall in a multi-class classification task (since the default binary variants are invalid).
This was addressed by #612.
Currently, we allow for all variants of the f1 metric in metrics.py, but we do not do the same for precision and recall.
This means that we cannot tune on precision/recall in a multi-class classification task (since the default binary variants are invalid).