I’m wondering if there is any way to normalize (based on training data) the probabilities that come out of a pipeline()?
Basically i’d like to try normalize my probabilities based on the training data such that i could say a probability of around 0.5 is about the average likelihood observed during training.
I was wondering if there might be any way to do that from within the pipeline() as opposed to having to keep track of whatever i need from the training stats and then running the last post processing step to normalize myself outside of the pipeline as obviously would be a little bit messier to maintain.