{"id":280,"date":"2017-04-05T13:04:33","date_gmt":"2017-04-05T12:04:33","guid":{"rendered":"http:\/\/pageperso.math.univ-toulouse.fr\/mllaw\/?page_id=280"},"modified":"2017-04-06T10:11:06","modified_gmt":"2017-04-06T09:11:06","slug":"apprentissage-debiaise","status":"publish","type":"page","link":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/home\/statisticien\/biais-dune-decision\/apprentissage-debiaise\/","title":{"rendered":"Apprentissage d\u00e9biais\u00e9"},"content":{"rendered":"<p>De nombreuses r\u00e9f\u00e9rences proposent d&#8217;adapter ou contraindre des m\u00e9thodes d&#8217;apprentissage \u00e0 construire une r\u00e8gle de d\u00e9cision sans biais ou de biais r\u00e9duit tout en pr\u00e9servant au mieux la qualit\u00e9 de pr\u00e9vision.<\/p>\n<ul>\n<li>Kamiran et al. (2010) proposent d&#8217;adapter la recherche des re\u00e8gles de division pour la d\u00e9finition des n\u0153uds d&#8217;un <a href=\"http:\/\/wikistat.fr\/pdf\/st-m-app-cart.pdf\">arbre de d\u00e9cision<\/a>.<\/li>\n<li>Calders et Verwer (2010) le font pour des classifieurs bay\u00e9siens na\u00effs.<\/li>\n<li>Tout r\u00e9cemment, Zafar et al. (2017) d\u00e9finissent des contraintes de &#8220;loyaut\u00e9&#8221; qui peuvent \u00eatre int\u00e9gr\u00e9es aux algorithmes d&#8217;estimation de m\u00e9thodes classiques\u00a0 comme la <a href=\"http:\/\/wikistat.fr\/pdf\/st-m-app-rlogit.pdf\">r\u00e9gression logistique<\/a> ou les\u00a0 <a href=\"http:\/\/wikistat.fr\/pdf\/st-m-app-svm.pdf\">SVM<\/a> tout en contr\u00f4lant la perte de pr\u00e9cision de la pr\u00e9vision.<\/li>\n<li>Dwrok et al. (2012) (?).<\/li>\n<li>Zemel et al. (2013) associent un individu \u00e0 une distribution dans un espace de repr\u00e9sentation, qui &#8220;oublie&#8221; l&#8217;appartenance \u00e0 un groupe prot\u00e9g\u00e9 tout en conservant le &#8220;plus &#8216;informations&#8221;. Il s&#8217;agit donc de la recherche d&#8217;une meilleure repr\u00e9sentation masquant l\u2019information sensible.<\/li>\n<li>Hajian et al. (2014) adoptent une autre strat\u00e9gie en d\u00e9biaisant une d\u00e9cision <em>apr\u00e8s<\/em> le processus d&#8217;apprentissage.<\/li>\n<\/ul>\n<h2>R\u00e9f\u00e9rences<\/h2>\n<p>T. Calders, S. VerwerThree (2010). <a href=\"https:\/\/pdfs.semanticscholar.org\/a087\/d3893af0276fe3b41924087670b03997f7af.pdf\">Naive Bayes Approaches for Discrimination-Free Classication<\/a> in Data Mining and Knowledge Discovery 21(2), 277\u2013292.<\/p>\n<p>C. Dwork, M. Hardt, T. Pitassi, O. Reingold, R. Zemel (2012). <a href=\"https:\/\/arxiv.org\/abs\/1104.3913\">Fairness Through Awareness<\/a> in ITCS, 214-226.<\/p>\n<p>S. Hajian, J. Domingo-Ferrer, A. Monreale, D. Pedreschi, F. Giannotti (2014). <a href=\"https:\/\/www.researchgate.net\/publication\/268746008_Discrimination-_and_privacy-aware_patterns\">Discrimination &#8211; and privacy &#8211; aware patterns<\/a> in Data Mining and Knowledge Discovery 29(6).<\/p>\n<p>F. Kamiran, T. Calders, M. Pechenizkiy (2010). <a href=\"http:\/\/wwwis.win.tue.nl\/~tcalders\/pubs\/ICDM2010KCP.pdf\">Discrimination Aware Decision Tree Learning<\/a> in ICDM, 869-874.<\/p>\n<p>M. B. Zafar, I. Valera, M. G. Rodriguez, K. P. Gummadi (2017). <a href=\"https:\/\/arxiv.org\/abs\/1507.05259\">Fairness Constraints: Mechanisms for Fair Classication<\/a> in International Conference on Articial Intelligence and Statistics (AISTATS), vol. 5.<\/p>\n<p>R. Zemel, Y. Wu, K. Swersky, T. Pitassi, C. Dwork (2013). <a href=\"https:\/\/www.cs.toronto.edu\/~toni\/Papers\/icml-final.pdf\">Learning Fair Representations<\/a> in\u00a0 JMLR W&amp;CP 28(3), 325\u2013333.<\/p>\n<p>&nbsp;<\/p>\n","protected":false},"excerpt":{"rendered":"<p>De nombreuses r\u00e9f\u00e9rences proposent d&#8217;adapter ou contraindre des m\u00e9thodes d&#8217;apprentissage \u00e0 construire une r\u00e8gle de d\u00e9cision sans biais ou de biais r\u00e9duit tout en pr\u00e9servant au mieux la qualit\u00e9 de pr\u00e9vision. Kamiran et al. (2010) proposent d&#8217;adapter la recherche des re\u00e8gles de division pour la d\u00e9finition des n\u0153uds d&#8217;un arbre de d\u00e9cision. Calders et Verwer [&hellip;]<\/p>\n","protected":false},"author":177,"featured_media":0,"parent":230,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"template-full-width.php","meta":{"footnotes":""},"class_list":["post-280","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/pages\/280","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/users\/177"}],"replies":[{"embeddable":true,"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/comments?post=280"}],"version-history":[{"count":4,"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/pages\/280\/revisions"}],"predecessor-version":[{"id":286,"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/pages\/280\/revisions\/286"}],"up":[{"embeddable":true,"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/pages\/230"}],"wp:attachment":[{"href":"https:\/\/perso.math.univ-toulouse.fr\/mllaw\/wp-json\/wp\/v2\/media?parent=280"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}