This is the jupyter lab project of Fair Logistic Regression (Fair-LR), a binary prediction model based on the Seldonian framework of Thomas et al., (2019). There are two notebook files: fair_lr and third_party_implementation. The fair_lr file is the main file of our LAK paper in 2021, where we implemented the fair logistics regression from our own understandings on Seldonian. The third_party_implementation includes a later implementation from a student, Abdul Hannan, from UMass, the institute where the authors of Seldonian Framework are from. Fair-LR allows researchers to define their own fairness evaluation metrics, utilize existing popular fairness metrics, and incorporate these metrics into model learning to learn to be fairer. Fair-LR has demonstrated capabilities to achieve both fair and accurate predictions.
Thomas, P. S., Castro da Silva, B., Barto, A. G., Giguere, S., Brun, Y., & Brunskill, E. (2019). Preventing undesirable behavior of intelligent machines. Science, 366(6468), 999-1004.