Fast Fair Regression via Efficient Approximations of Mutual Information

Most work in algorithmic fairness to date has focused on discrete outcomes, such as deciding whether to grant someone a loan or not. In these classification settings, group fairness criteria such as independence, separation and sufficiency can be measured directly by comparing rates of outcomes betw...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Steinberg, Daniel, Reid, Alistair, O'Callaghan, Simon, Lattimore, Finnian, McCalman, Lachlan, Caetano, Tiberio
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Steinberg, Daniel
Reid, Alistair
O'Callaghan, Simon
Lattimore, Finnian
McCalman, Lachlan
Caetano, Tiberio
description Most work in algorithmic fairness to date has focused on discrete outcomes, such as deciding whether to grant someone a loan or not. In these classification settings, group fairness criteria such as independence, separation and sufficiency can be measured directly by comparing rates of outcomes between subpopulations. Many important problems however require the prediction of a real-valued outcome, such as a risk score or insurance premium. In such regression settings, measuring group fairness criteria is computationally challenging, as it requires estimating information-theoretic divergences between conditional probability density functions. This paper introduces fast approximations of the independence, separation and sufficiency group fairness criteria for regression models from their (conditional) mutual information definitions, and uses such approximations as regularisers to enforce fairness within a regularised risk minimisation framework. Experiments in real-world datasets indicate that in spite of its superior computational efficiency our algorithm still displays state-of-the-art accuracy/fairness tradeoffs.
doi_str_mv 10.48550/arxiv.2002.06200
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2002_06200</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2002_06200</sourcerecordid><originalsourceid>FETCH-LOGICAL-a670-1adf50c348fb1c8dd389960a37471d98c1a68940a77a54f1050c36fdca09ed003</originalsourceid><addsrcrecordid>eNotj01rAjEYhHPpoWh_QE_mD-z2jcnm4yjiqmARxPvyNh8loLtLsor9967a08DMMMxDyCeDUuiqgi9Mt3gt5wDzEuQo72RdYx5ojTHRg_9NPufYtfQaka5CiDb6dqCLvk_dLZ5xGLNMu0C_L8MFT3Tbhi697Cl5C3jK_uNfJ-RYr47LTbHbr7fLxa5AqaBg6EIFlgsdfpjVznFtjATkSijmjLYMpTYCUCmsRGDwKMvgLILxDoBPyOw1-yRp-jS-Sn_Ng6h5EvE7QWhGHg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Fast Fair Regression via Efficient Approximations of Mutual Information</title><source>arXiv.org</source><creator>Steinberg, Daniel ; Reid, Alistair ; O'Callaghan, Simon ; Lattimore, Finnian ; McCalman, Lachlan ; Caetano, Tiberio</creator><creatorcontrib>Steinberg, Daniel ; Reid, Alistair ; O'Callaghan, Simon ; Lattimore, Finnian ; McCalman, Lachlan ; Caetano, Tiberio</creatorcontrib><description>Most work in algorithmic fairness to date has focused on discrete outcomes, such as deciding whether to grant someone a loan or not. In these classification settings, group fairness criteria such as independence, separation and sufficiency can be measured directly by comparing rates of outcomes between subpopulations. Many important problems however require the prediction of a real-valued outcome, such as a risk score or insurance premium. In such regression settings, measuring group fairness criteria is computationally challenging, as it requires estimating information-theoretic divergences between conditional probability density functions. This paper introduces fast approximations of the independence, separation and sufficiency group fairness criteria for regression models from their (conditional) mutual information definitions, and uses such approximations as regularisers to enforce fairness within a regularised risk minimisation framework. Experiments in real-world datasets indicate that in spite of its superior computational efficiency our algorithm still displays state-of-the-art accuracy/fairness tradeoffs.</description><identifier>DOI: 10.48550/arxiv.2002.06200</identifier><language>eng</language><subject>Computer Science - Learning ; Statistics - Machine Learning</subject><creationdate>2020-02</creationdate><rights>http://arxiv.org/licenses/nonexclusive-distrib/1.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,776,881</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2002.06200$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2002.06200$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Steinberg, Daniel</creatorcontrib><creatorcontrib>Reid, Alistair</creatorcontrib><creatorcontrib>O'Callaghan, Simon</creatorcontrib><creatorcontrib>Lattimore, Finnian</creatorcontrib><creatorcontrib>McCalman, Lachlan</creatorcontrib><creatorcontrib>Caetano, Tiberio</creatorcontrib><title>Fast Fair Regression via Efficient Approximations of Mutual Information</title><description>Most work in algorithmic fairness to date has focused on discrete outcomes, such as deciding whether to grant someone a loan or not. In these classification settings, group fairness criteria such as independence, separation and sufficiency can be measured directly by comparing rates of outcomes between subpopulations. Many important problems however require the prediction of a real-valued outcome, such as a risk score or insurance premium. In such regression settings, measuring group fairness criteria is computationally challenging, as it requires estimating information-theoretic divergences between conditional probability density functions. This paper introduces fast approximations of the independence, separation and sufficiency group fairness criteria for regression models from their (conditional) mutual information definitions, and uses such approximations as regularisers to enforce fairness within a regularised risk minimisation framework. Experiments in real-world datasets indicate that in spite of its superior computational efficiency our algorithm still displays state-of-the-art accuracy/fairness tradeoffs.</description><subject>Computer Science - Learning</subject><subject>Statistics - Machine Learning</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2020</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj01rAjEYhHPpoWh_QE_mD-z2jcnm4yjiqmARxPvyNh8loLtLsor9967a08DMMMxDyCeDUuiqgi9Mt3gt5wDzEuQo72RdYx5ojTHRg_9NPufYtfQaka5CiDb6dqCLvk_dLZ5xGLNMu0C_L8MFT3Tbhi697Cl5C3jK_uNfJ-RYr47LTbHbr7fLxa5AqaBg6EIFlgsdfpjVznFtjATkSijmjLYMpTYCUCmsRGDwKMvgLILxDoBPyOw1-yRp-jS-Sn_Ng6h5EvE7QWhGHg</recordid><startdate>20200214</startdate><enddate>20200214</enddate><creator>Steinberg, Daniel</creator><creator>Reid, Alistair</creator><creator>O'Callaghan, Simon</creator><creator>Lattimore, Finnian</creator><creator>McCalman, Lachlan</creator><creator>Caetano, Tiberio</creator><scope>AKY</scope><scope>EPD</scope><scope>GOX</scope></search><sort><creationdate>20200214</creationdate><title>Fast Fair Regression via Efficient Approximations of Mutual Information</title><author>Steinberg, Daniel ; Reid, Alistair ; O'Callaghan, Simon ; Lattimore, Finnian ; McCalman, Lachlan ; Caetano, Tiberio</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a670-1adf50c348fb1c8dd389960a37471d98c1a68940a77a54f1050c36fdca09ed003</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2020</creationdate><topic>Computer Science - Learning</topic><topic>Statistics - Machine Learning</topic><toplevel>online_resources</toplevel><creatorcontrib>Steinberg, Daniel</creatorcontrib><creatorcontrib>Reid, Alistair</creatorcontrib><creatorcontrib>O'Callaghan, Simon</creatorcontrib><creatorcontrib>Lattimore, Finnian</creatorcontrib><creatorcontrib>McCalman, Lachlan</creatorcontrib><creatorcontrib>Caetano, Tiberio</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv Statistics</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Steinberg, Daniel</au><au>Reid, Alistair</au><au>O'Callaghan, Simon</au><au>Lattimore, Finnian</au><au>McCalman, Lachlan</au><au>Caetano, Tiberio</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Fast Fair Regression via Efficient Approximations of Mutual Information</atitle><date>2020-02-14</date><risdate>2020</risdate><abstract>Most work in algorithmic fairness to date has focused on discrete outcomes, such as deciding whether to grant someone a loan or not. In these classification settings, group fairness criteria such as independence, separation and sufficiency can be measured directly by comparing rates of outcomes between subpopulations. Many important problems however require the prediction of a real-valued outcome, such as a risk score or insurance premium. In such regression settings, measuring group fairness criteria is computationally challenging, as it requires estimating information-theoretic divergences between conditional probability density functions. This paper introduces fast approximations of the independence, separation and sufficiency group fairness criteria for regression models from their (conditional) mutual information definitions, and uses such approximations as regularisers to enforce fairness within a regularised risk minimisation framework. Experiments in real-world datasets indicate that in spite of its superior computational efficiency our algorithm still displays state-of-the-art accuracy/fairness tradeoffs.</abstract><doi>10.48550/arxiv.2002.06200</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2002.06200
ispartof
issn
language eng
recordid cdi_arxiv_primary_2002_06200
source arXiv.org
subjects Computer Science - Learning
Statistics - Machine Learning
title Fast Fair Regression via Efficient Approximations of Mutual Information
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-05T14%3A04%3A52IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Fast%20Fair%20Regression%20via%20Efficient%20Approximations%20of%20Mutual%20Information&rft.au=Steinberg,%20Daniel&rft.date=2020-02-14&rft_id=info:doi/10.48550/arxiv.2002.06200&rft_dat=%3Carxiv_GOX%3E2002_06200%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true