Open
Description
In case the ground truth for the model is not available it should still be possible to run aequitas to get the metrics that do not depend on this input. Right now the Group() class requires label_value as input but it's not needed for metrics such as predictive positive ratio. One can input label_value with eg. 0 or 1 but then some supervised metrics will still be calculated incorrectly based on this input which can be confusing.
Metadata
Metadata
Assignees
Labels
No labels