Joint Transfer of Model Knowledge and Fairness Over Domains Using Wasserstein Distance
Owing to the increasing use of machine learning in our daily lives, the problem of fairness has recently become an important topic in machine learning societies. Recent studies regarding fairness in machine learning have been conducted to attempt to ensure statistical independence between individual...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2020-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/9129731/ |
id |
doaj-930e7265724a47a7889b128c92587d84 |
---|---|
record_format |
Article |
spelling |
doaj-930e7265724a47a7889b128c92587d842021-03-30T01:55:14ZengIEEEIEEE Access2169-35362020-01-01812378312379810.1109/ACCESS.2020.30059879129731Joint Transfer of Model Knowledge and Fairness Over Domains Using Wasserstein DistanceTaeho Yoon0https://orcid.org/0000-0003-0508-0081Jaewook Lee1https://orcid.org/0000-0001-5720-8337Woojin Lee2https://orcid.org/0000-0002-7432-4185Department of Mathematical Sciences, Seoul National University, Seoul, South KoreaDepartment of Industrial Engineering, Seoul National University, Seoul, South KoreaDepartment of Mathematical Sciences, Seoul National University, Seoul, South KoreaOwing to the increasing use of machine learning in our daily lives, the problem of fairness has recently become an important topic in machine learning societies. Recent studies regarding fairness in machine learning have been conducted to attempt to ensure statistical independence between individual model predictions and designated sensitive attributes. However, in reality, cases exist in which the sensitive variables of data used for learning models differ from the data upon which the model is applied. In this paper, we investigate a methodology for developing a fair classification model for data with limited or no labels, by transferring knowledge from another data domain where information is fully available. This is done by controlling the Wasserstein distances between relevant distributions. Subsequently, we obtain a fair model that could be successfully applied to two datasets with different sensitive attributes. We present theoretical results validating that our approach provably transfers both classification performance and fairness over domains. Experimental results show that our method does indeed promote fairness for the target domain, while retaining reasonable classification accuracy, and that it often outperforms comparative models in terms of joint fairness.https://ieeexplore.ieee.org/document/9129731/Fair machine learningfair classificationdemographic parityequal opportunitydomain adaptationtransfer learning |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Taeho Yoon Jaewook Lee Woojin Lee |
spellingShingle |
Taeho Yoon Jaewook Lee Woojin Lee Joint Transfer of Model Knowledge and Fairness Over Domains Using Wasserstein Distance IEEE Access Fair machine learning fair classification demographic parity equal opportunity domain adaptation transfer learning |
author_facet |
Taeho Yoon Jaewook Lee Woojin Lee |
author_sort |
Taeho Yoon |
title |
Joint Transfer of Model Knowledge and Fairness Over Domains Using Wasserstein Distance |
title_short |
Joint Transfer of Model Knowledge and Fairness Over Domains Using Wasserstein Distance |
title_full |
Joint Transfer of Model Knowledge and Fairness Over Domains Using Wasserstein Distance |
title_fullStr |
Joint Transfer of Model Knowledge and Fairness Over Domains Using Wasserstein Distance |
title_full_unstemmed |
Joint Transfer of Model Knowledge and Fairness Over Domains Using Wasserstein Distance |
title_sort |
joint transfer of model knowledge and fairness over domains using wasserstein distance |
publisher |
IEEE |
series |
IEEE Access |
issn |
2169-3536 |
publishDate |
2020-01-01 |
description |
Owing to the increasing use of machine learning in our daily lives, the problem of fairness has recently become an important topic in machine learning societies. Recent studies regarding fairness in machine learning have been conducted to attempt to ensure statistical independence between individual model predictions and designated sensitive attributes. However, in reality, cases exist in which the sensitive variables of data used for learning models differ from the data upon which the model is applied. In this paper, we investigate a methodology for developing a fair classification model for data with limited or no labels, by transferring knowledge from another data domain where information is fully available. This is done by controlling the Wasserstein distances between relevant distributions. Subsequently, we obtain a fair model that could be successfully applied to two datasets with different sensitive attributes. We present theoretical results validating that our approach provably transfers both classification performance and fairness over domains. Experimental results show that our method does indeed promote fairness for the target domain, while retaining reasonable classification accuracy, and that it often outperforms comparative models in terms of joint fairness. |
topic |
Fair machine learning fair classification demographic parity equal opportunity domain adaptation transfer learning |
url |
https://ieeexplore.ieee.org/document/9129731/ |
work_keys_str_mv |
AT taehoyoon jointtransferofmodelknowledgeandfairnessoverdomainsusingwassersteindistance AT jaewooklee jointtransferofmodelknowledgeandfairnessoverdomainsusingwassersteindistance AT woojinlee jointtransferofmodelknowledgeandfairnessoverdomainsusingwassersteindistance |
_version_ |
1724186237888102400 |