"Are Two Heads the Same as One? Identifying Disparate Treatment in Fair Neural Networks".

Welcome. This is the code release of our 2022 NeurIPS paper Are Two Heads the Same as One? Identifying Disparate Treatment in Fair Neural Networks.

Examples

Please read the overview_notebook.ipynb with examples, results, and explanations. It takes some time to run all models, but if you want to run all models yourself, run all_jobs.sh after the following setup.

Setup

1. Download img_align_celeba.zip and list_attr_celeba.txt from http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html and move it to data/celeba/.

2. Create a virtual environment with python=3.8, activate it, and run

python -m pip install -r requirements.txt

3. Install your torch and torchvision. We used torch==1.12.1+cu116 and torchvision==0.13.1+cu116.

pip install torch==1.12.1+cu116 torchvision==0.13.1+cu116 --extra-index-url https://download.pytorch.org/whl/cu116

Citing this paper

If you use this software or want to refer to it, please cite the following publication:

@inproceedings{lohaus2022,
  title={Are Two Heads the Same as One? Identifying Disparate Treatment in Fair Neural Networks},
  author={Lohaus, Michael and Kleindessner, Matth\"aus and Kenthapadi, Krishnaram and Locatello, Francesco and Russell, Chris},
  booktitle={Neural Information Processing Systems (NeurIPS)},
  year={2022}
}