TianxiangZhao / fairlearn Public

Notifications You must be signed in to change notification settings
Fork 3
Star 4

Official code for WSDM 2022 paper: Towards Fair Classifiers Without Sensitive Attributes: Exploring Biases in Related Featuresto be added

4 stars 3 forks Branches Tags Activity

Notifications

Name		Name	Last commit message	Last commit date
Latest commit History 11 Commits
data		data
README.md		README.md
main.py		main.py
utils.py		utils.py

Repository files navigation

fair learning without sensitive attributes

Official code for WSDM 2022 paper: Towards Fair Classifiers Without Sensitive Attributes: Exploring Biases in Related Features

Dataset

Two datasets, Law_school and Compas, are provided in this project.

Algorithm

We provide several algorithms for fair learning:

corre: constrain correlation with sensitive attributes
groupTPR: regularize group-wise true positive rate for fairness
remove: remove related attributes
learnCorre: learn to constrain correlation with related attributes

Example

An example on adult dataset is provided here:

python main.py --method="learnCorre" --dataset=adult --related age --r_weight 0.1 --weightSum=0.1 --beta=0.4 --seed=42

About

Official code for WSDM 2022 paper: Towards Fair Classifiers Without Sensitive Attributes: Exploring Biases in Related Featuresto be added

Report repository

Releases

No releases published

Packages

No packages published

Languages

Python 100.0%