Skip to content

Latest commit

 

History

History
62 lines (49 loc) · 2.01 KB

README.md

File metadata and controls

62 lines (49 loc) · 2.01 KB
Raw

On the Within-Group Fairness of Screening Classifiers

This is a repository containing code for the paper:

N. Okati, S. Tsirtsis, and M. Gomez Rodriguez. On the Within-Group Discrimination of Screening Classifiers., ICML 2023.

The paper is available here.

Pre-requisites

To install all the requirements, on a machine with conda installed, run

conda env create -f environment.yml
source activate wgm

Preparation

The parameters required for every set of experiments are in ./scripts/params_*.py files. The current settings are those used in our experiments. The first time that you run the experiments the data will be downloaded and saved in the data folder.

If you are using a machine with Slurm workload manager set submit = True in ./scripts/params_*.py. You can then increase the n_runs parameter which specifies the number of runs. We used 100 in our experiments. Set submit = False if you run the experiments on your local machine and make sure the number of runs is small.

Execution

python ./scripts/run_exp_bins.py

Generate Figures

To generate Figures 1, 5, 9

python ./scripts/plot_exp_violations.py

To generate Figures 2, 7

python ./scripts/plot_discrimination.py
python ./scripts/plot_exp_group_discrimination.py

To generate Figures 3, 4, 8

python ./scripts/plot_exp_bins.py

To generate Figures 6

python ./scripts/plot_wgc_eps.py

Citation

If you use parts of the code in this repository for your own research, please consider citing:

@inproceedings{okati2023on,
        title={On the Within-Group Discrimination of Screening Classifiers},
        author={Okati, Nastaran and Tsirtsis, Stratis and Gomez-Rodriguez, Manuel},
        booktitle={Proceedings of the 40th International Conference on Machine Learning},
        year={2023}
}