v0.11.0#

This release greately improves upon the scikit-learn compatibility of the AdversarialFairnessClassifier and AdversarialFairnessRegressor estimators. It also includes a number of other improvements.

scikit-learn compatibility#

  • The estimators AdversarialFairnessClassifier and AdversarialFairnessRegressor now pass all but one scikit-learn estimator checks. Changes in #1410, #1405, #1399, #1398, #1394, #1393, #1369, and the ones included below in this section.

  • Added classes_ attribute to AdversarialFairnessClassifier and AdversarialFairnessRegressor in #1384.

  • Removed _AdversarialFairness.decision_function because it did not follow the expected contract with _AdversarialFairness.predict in #1397.

Other improvements#

  • Added disparity metric equal_opportunity_{difference,ratio} as a frequently used alias for true_positive_rate_{difference,ratio}.

  • Added numpy 2.0.0 compatibility in #1387.

  • Fixed pandas warnings in #1363 and #1339.

  • Fixed output dimensions bug in FloatingPointTransformer in #1333.