This repository is meant to facilitate the benchmarking of fairness aware machine learning algorithms.
The associated paper is:
A comparative study of fairness-enhancing interventions in machine learning by Sorelle A. Friedler, Carlos Scheidegger, Suresh Venkatasubramanian, Sonam Choudhary, Evan P. Hamilton, and Derek Roth. https://arxiv.org/abs/1802.04422
To install this software run:
pip3 install fairness
The below instructions are still in the process of being updated to work with the new pip install-able version.
To run the benchmarks:
$ from fairness.benchmark import run $ run()
This will write out metrics for each dataset to the results/ directory.
To generate graphs and other analysis run:
If you do not yet have all the packages installed, you may need to run:
pip install -r requirements.txt
Optional: The benchmarks rely on preprocessed versions of the datasets that have been included in the repository. If you would like to regenerate this preprocessing, run the below command before running the benchmark script:
To add new datasets or algorithms, see the instructions in the readme files in those directories.
(We tested on Ubuntu 16.04, your mileage may vary)
$ sudo apt-get install python3-dev
To regenerate figures (this is messy right now. we're working on it)
Python requirements (use pip):
brew install pandocon a Mac or
apt-get install pandocon Linux)
R package requirements (use