This repository provides an implementation of fairness-aware regularization for notions of fairness based on (conditional) independence. This code was used to run the experiments of the paper:
First, you can create two conda environments from the YML files, one for the library facl only and one for library and notebooks. If you wish to run the examples, you should choose 'env_library_and_notebooks.yml'.
conda env create -f env_library_and_notebooks.yml
conda activate continuous-fairness-all
You can find several examples of use in the directory examples
in the form of
jupyter notebooks.
The license can be found on in the file LICENSE.
If you use this code please cite
@InProceedings{pmlr-v97-mary19a,
title = {Fairness-Aware Learning for Continuous Attributes and Treatments},
author = {Mary, Jeremie and Calauz{\`e}nes, Cl{\'e}ment and El Karoui, Noureddine},
booktitle = {Proceedings of the 36th International Conference on Machine Learning},
pages = {4382--4391},
year = {2019},
editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan},
volume = {97},
series = {Proceedings of Machine Learning Research},
address = {Long Beach, California, USA},
month = {09--15 Jun},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v97/mary19a/mary19a.pdf},
url = {http://proceedings.mlr.press/v97/mary19a.html},
abstract = {We address the problem of algorithmic fairness: ensuring that the outcome of a classifier is not biased towards certain values of sensitive variables such as age, race or gender. As common fairness metrics can be expressed as measures of (conditional) independence between variables, we propose to use the Rényi maximum correlation coefficient to generalize fairness measurement to continuous variables. We exploit Witsenhausen’s characterization of the Rényi correlation coefficient to propose a differentiable implementation linked to $f$-divergences. This allows us to generalize fairness-aware learning to continuous variables by using a penalty that upper bounds this coefficient. Theses allows fairness to be extented to variables such as mixed ethnic groups or financial status without thresholds effects. This penalty can be estimated on mini-batches allowing to use deep nets. Experiments show favorable comparisons to state of the art on binary variables and prove the ability to protect continuous ones}
}