Skip to content

Code for our paper "Fairwashing: the risk of rationalization" (https://arxiv.org/abs/1901.09749) accepted at ICML 2019.

License

Notifications You must be signed in to change notification settings

aivodji/LaundryML

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

7 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

LaundryML

**LaundryML is a regularized model enumeration algorithm that enumerates rule lists for rationalization.

This code is based on the implementation of the Certifiably Optimal RulE ListS (CORELS) algorithm. For information on CORELS, please visit its website.

Python dependencies

  • pip install -r requirements.txt

C/C++ dependencies

  • gmp (GNU Multiple Precision Arithmetic Library)
  • mpfr (GNU MPFR Library for multiple-precision floating-point computations; depends on gmp)
  • libmpc (GNU MPC for arbitrarily high precision and correct rounding; depends on gmp and mpfr)

Compilation

  • make

Scripts for model rationalization

  • [Adult Income]: python benchmark_adult.py
  • [ProPublica Recidivism]: python benchmark_compas.py
  • Result will be generated in the folder "output"

Scripts for outcome rationalization

  • [Adult Income]: python benchmark_local_adult.py
  • [ProPublica Recidivism]: python benchmark_local_compas.py
  • Result will be generated in the folder "res_local"

Visualizing results for model rationalization

Models performances

  • First run python summary_rationalization_both.py
  • Run Rscript both_analysis.R in the "plotting_scripts" repository
  • Results are saved in plotting_scripts/data/graphs

Audit of best models

  • First run python selection_adult.py and python selection_compas.py to get the best models='s id, beta and lambda
  • Run python summary_audit_adult.py and python summary_compas_adult.py with the corresponding pareameters
  • Results are saved in plotting_scripts/data/graphs

Visualizing results for outcome rationalization

Models performances

  • Run Rscript analysis_local_adult.R and Rscript analysis_local_adult.R in the "plotting_scripts" repository
  • Results are saved in plotting_scripts/data/graphs

Citing this work

@InProceedings{pmlr-v97-aivodji19a,
  title = 	 {Fairwashing: the risk of rationalization},
  author = 	 {Aivodji, Ulrich and Arai, Hiromi and Fortineau, Olivier and Gambs, S{\'e}bastien and Hara, Satoshi and Tapp, Alain},
  booktitle = 	 {Proceedings of the 36th International Conference on Machine Learning},
  pages = 	 {161--170},
  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/aivodji19a/aivodji19a.pdf},
  url = 	 {http://proceedings.mlr.press/v97/aivodji19a.html},
}

About

Code for our paper "Fairwashing: the risk of rationalization" (https://arxiv.org/abs/1901.09749) accepted at ICML 2019.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published