Skip to content

Code for "Fair Densities via Boosting the Sufficient Statistics of Exponential Families"

Notifications You must be signed in to change notification settings

alexandersoen/fbde

Repository files navigation

Code SI

General code for "Fair Densities via Boosting the Sufficient Statistics of Exponential Families".

Installation

We configure using conda environment.yml provides dependencies to run the code. Additional files need to be setup and downloaded for AIF360 datasets to work.

Demo

run_discrete.py runs FBDE for discrete datasets. An example:

ipython run_discrete.py -- --equal-rr --boosting-steps 32 --max-depth 8 --sattr 'sex' --seed 1 --dataset 'compas' --sr 0.8 --leverage 'exact'

Alternatively, one can run the continuous dataset code with run_continuous.py:

ipython run_continuous.py -- --equal-rr --boosting-steps 8 --max-depth 8  --seed 1 --dataset 'minneapolis' --sr 0.8 --leverage 'exact'

Additionally, we provide *_data.py scripts to run evaluation on the raw data.

Datasets

One can use compas and adult for --dataset options, with sex or race as --sattr. Otherwise, one can use german and dutch for --dataset options, with sex or age as --sattr.

Citation

Fair Densities via Boosting the Sufficient Statistics of Exponential Families
Alexander Soen, Hisham Husain, Richard Nock
International Conference on Machine Learning, ICML 2023

Please cite the corresponding paper when using the code

About

Code for "Fair Densities via Boosting the Sufficient Statistics of Exponential Families"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages