* changed how aggregation happens in policy ensembles. Added policy learning to mulit-investment attribution. Fixed some mistkaes in that case study
* added clipping to the denominator in the dr correction to avoid division by zero, in the dr estimate of ate and att in cfdml.
* added policy learning module
* added cython policy tree and policy forest
* extended policy cate interpreter to interpret multiple treatments using the new policy tree
* added doubly robust policy learning tree and doubly robust policy learning forest
* fixed randomness in weightedkfold, that was causing tests to fail due to non-fixed-randomness behavior
* added notebook on policy learning
pre-commit-ci[bot]
Keith Battocchi
OliverTseng
vsyrgkanis