Optimized Score Transformation for Fair Classification

Dennis Wei, Karthikeyan Natesan Ramamurthy, Flavio Calmon
Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, PMLR 108:1673-1683, 2020.

Abstract

This paper considers fair probabilistic classification where the outputs of primary interest are predicted probabilities, commonly referred to as scores. We formulate the problem of transforming scores to satisfy fairness constraints while minimizing the loss in utility. The formulation can be applied either to post-process classifier outputs or to pre-process training data, thus allowing maximum freedom in selecting a classification algorithm. We derive a closed-form expression for the optimal transformed scores and a convex optimization problem for the transformation parameters. In the population limit, the transformed score function is the fairness-constrained minimizer of cross-entropy with respect to the optimal unconstrained scores. In the finite sample setting, we propose to approach this solution using a combination of standard probabilistic classifiers and ADMM. Comprehensive experiments comparing to 10 existing methods show that the proposed FairScoreTransformer has advantages for score-based metrics such as Brier score and AUC while remaining competitive for binary label-based metrics such as accuracy.

Cite this Paper


BibTeX
@InProceedings{pmlr-v108-wei20a, title = {Optimized Score Transformation for Fair Classification}, author = {Wei, Dennis and Ramamurthy, Karthikeyan Natesan and Calmon, Flavio}, booktitle = {Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics}, pages = {1673--1683}, year = {2020}, editor = {Chiappa, Silvia and Calandra, Roberto}, volume = {108}, series = {Proceedings of Machine Learning Research}, month = {26--28 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v108/wei20a/wei20a.pdf}, url = {https://proceedings.mlr.press/v108/wei20a.html}, abstract = {This paper considers fair probabilistic classification where the outputs of primary interest are predicted probabilities, commonly referred to as scores. We formulate the problem of transforming scores to satisfy fairness constraints while minimizing the loss in utility. The formulation can be applied either to post-process classifier outputs or to pre-process training data, thus allowing maximum freedom in selecting a classification algorithm. We derive a closed-form expression for the optimal transformed scores and a convex optimization problem for the transformation parameters. In the population limit, the transformed score function is the fairness-constrained minimizer of cross-entropy with respect to the optimal unconstrained scores. In the finite sample setting, we propose to approach this solution using a combination of standard probabilistic classifiers and ADMM. Comprehensive experiments comparing to 10 existing methods show that the proposed FairScoreTransformer has advantages for score-based metrics such as Brier score and AUC while remaining competitive for binary label-based metrics such as accuracy.} }
Endnote
%0 Conference Paper %T Optimized Score Transformation for Fair Classification %A Dennis Wei %A Karthikeyan Natesan Ramamurthy %A Flavio Calmon %B Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2020 %E Silvia Chiappa %E Roberto Calandra %F pmlr-v108-wei20a %I PMLR %P 1673--1683 %U https://proceedings.mlr.press/v108/wei20a.html %V 108 %X This paper considers fair probabilistic classification where the outputs of primary interest are predicted probabilities, commonly referred to as scores. We formulate the problem of transforming scores to satisfy fairness constraints while minimizing the loss in utility. The formulation can be applied either to post-process classifier outputs or to pre-process training data, thus allowing maximum freedom in selecting a classification algorithm. We derive a closed-form expression for the optimal transformed scores and a convex optimization problem for the transformation parameters. In the population limit, the transformed score function is the fairness-constrained minimizer of cross-entropy with respect to the optimal unconstrained scores. In the finite sample setting, we propose to approach this solution using a combination of standard probabilistic classifiers and ADMM. Comprehensive experiments comparing to 10 existing methods show that the proposed FairScoreTransformer has advantages for score-based metrics such as Brier score and AUC while remaining competitive for binary label-based metrics such as accuracy.
APA
Wei, D., Ramamurthy, K.N. & Calmon, F.. (2020). Optimized Score Transformation for Fair Classification. Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 108:1673-1683 Available from https://proceedings.mlr.press/v108/wei20a.html.

Related Material