We present a systematic approach for achieving fairness in a binary
classification setting. While we focus on two well-known quantitative
definitions of fairness, our approach encompasses many other previously studied
definitions as special cases. The key idea is to reduce fair classification to
a sequence of cost-sensitive classification problems, whose solutions yield a
randomized classifier with the lowest (empirical) error subject to the desired
constraints. We introduce two reductions that work for any representation of
the cost-sensitive classifier and compare favorably to prior baselines on a
variety of data sets, while overcoming several of their disadvantages.
Description
[1803.02453] A Reductions Approach to Fair Classification
%0 Generic
%1 agarwal2018reductions
%A Agarwal, Alekh
%A Beygelzimer, Alina
%A Dudík, Miroslav
%A Langford, John
%A Wallach, Hanna M.
%D 2018
%K classification
%T A Reductions Approach to Fair Classification.
%U http://arxiv.org/abs/1803.02453
%X We present a systematic approach for achieving fairness in a binary
classification setting. While we focus on two well-known quantitative
definitions of fairness, our approach encompasses many other previously studied
definitions as special cases. The key idea is to reduce fair classification to
a sequence of cost-sensitive classification problems, whose solutions yield a
randomized classifier with the lowest (empirical) error subject to the desired
constraints. We introduce two reductions that work for any representation of
the cost-sensitive classifier and compare favorably to prior baselines on a
variety of data sets, while overcoming several of their disadvantages.
@misc{agarwal2018reductions,
abstract = {We present a systematic approach for achieving fairness in a binary
classification setting. While we focus on two well-known quantitative
definitions of fairness, our approach encompasses many other previously studied
definitions as special cases. The key idea is to reduce fair classification to
a sequence of cost-sensitive classification problems, whose solutions yield a
randomized classifier with the lowest (empirical) error subject to the desired
constraints. We introduce two reductions that work for any representation of
the cost-sensitive classifier and compare favorably to prior baselines on a
variety of data sets, while overcoming several of their disadvantages.},
added-at = {2018-12-07T10:20:56.000+0100},
author = {Agarwal, Alekh and Beygelzimer, Alina and Dudík, Miroslav and Langford, John and Wallach, Hanna M.},
biburl = {https://www.bibsonomy.org/bibtex/23fbc485ef7aff66829fc068f88027c8e/jpvaldes},
description = {[1803.02453] A Reductions Approach to Fair Classification},
interhash = {b417a9296ee93e7076a5e5e8dcb288ab},
intrahash = {3fbc485ef7aff66829fc068f88027c8e},
keywords = {classification},
note = {cite arxiv:1803.02453},
timestamp = {2018-12-07T10:20:56.000+0100},
title = {A Reductions Approach to Fair Classification.},
url = {http://arxiv.org/abs/1803.02453},
year = 2018
}