Moving metrics definition in simple module
we are having lots of metrics in the meanwhile, eff. vs rej. etc.
they are mostly embedded in high level functions which directly plot etc., it would be good to have a simple api to call these kind of metrics.
There are several approaches, either we add a small module in umami, e.g. umami.metrics
, we do it in a full new repository, or we add it in a bigger framework hep-ml
, scikit-learn
Idea
A good first step would be to move the metrics we are using in a simple stand-alone umami module e.g. metrics
and in a second step we could think of adding it to an existing python package (e.g. hep_ml
). This way we would already have most of the implementation.
The following metrics are needed
-
background rejection for specific signal (b) efficiency (one point in ROC) -
background rejection as function of signal (b) efficiency (ROC) -
background rejection as function of arbitrary variable signal (b) efficiency -
cumulative efficiency as a function of b-tagging discriminant (Figure 7) -
fc scan ROCs slide 9/10
would be good to have some opinions here how we could best do that