Mean Parity Fair Regression in RKHS

The official implementation of the paper Mean Parity Fair Regression in RKHS. This implementation is built upon https://github.com/steven7woo/fair_regression_reduction.

In this fair_reg.py, we implement three methods used in the paper including:

• MP_Fair_regression: Our method for achieving MP fairness or Covariance-based fairness
• MP_Penalty_regression: A penalty-based method for achieving MP fairness or Covariance-based fairness
• Fair_kernel_learning: A penalty-based method for achieving Covariance-based fairness

Steps to run the code:

1. Download the preprocessed dataset from https://github.com/steven7woo/fair_regression_reduction by running the following code in your terminal

   bash collect_data.sh

2. Install the dependencies by running the following code in your terminal

   pip install -r requirements.txt

3. Run our demo notebook in your Jupyter Notebook Environment.

For other baselines used in this paper, we refer users to their official implementations:

• Nonconvex Regression with Fairness Constraints: https://github.com/jkomiyama/fairregresion
• Reduction Based Algorithm: https://github.com/steven7woo/fair_regression_reduction

If you have any problem, please contact Shaokui Wei.

If you find this repo useful, consider citing our paper

  @inproceedings{wei2023mean,
  title={Mean Parity Fair Regression in RKHS},
  author={Wei, Shaokui and Liu, Jiayin and Li, Bing and Zha, Hongyuan},
  booktitle={International Conference on Artificial Intelligence and Statistics},
  pages={4602--4628},
  year={2023},
  organization={PMLR}
}