Fair Regression: Reduction-Based Algorithms

Implementation for a reduction-based algorithm for fair regression subject to the constraint of demographic parity (also called statistical parity).

If you find thie repository useful for your research, please consider citing our work:

@inproceedings{ADW19,
  author    = {Alekh Agarwal and
               Miroslav Dud{\'{\i}}k and
               Zhiwei Steven Wu},
  title     = {Fair Regression: Quantitative Definitions and Reduction-Based Algorithms},
  booktitle = {Proceedings of the 36th International Conference on Machine Learning,
               {ICML} 2019, 9-15 June 2019, Long Beach, California, {USA}},
  year      = {2019},
  url       = {http://proceedings.mlr.press/v97/agarwal19d.html}
}

arXiv link to this paper

Requirements

To run the code the following packages need to be installed:

• Gurobi solver.
• Python package gurobipy. Avaiable with Anaconda.
• Python XGB package. Installation guide.

Dataset

We include three datasets.

• Adult Income
• LSAC National Longitudinal (Law School)
• Communities and Crime

Usage

• To train a fair regression model, run exp_grad.py.
• Run run_exp.py to reproduce results in the paper.

Bounded group loss

This implementation focuses on demographic parity. For fair regression with bounded group loss constraint, please see the implementation in fairlearn library.