{"id":3384,"date":"2020-06-08T16:14:09","date_gmt":"2020-06-08T16:14:09","guid":{"rendered":"https:\/\/datascientest.com\/?p=3384"},"modified":"2023-11-09T14:14:17","modified_gmt":"2023-11-09T14:14:17","slug":"peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime","status":"publish","type":"post","link":"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime","title":{"rendered":"Peut-on avoir confiance aux pr\u00e9dictions d\u2019une IA : L\u2019approche LIME"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"3384\" class=\"elementor elementor-3384\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-94ad529 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"94ad529\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-b7450cc\" data-id=\"b7450cc\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-7e01ecd elementor-widget elementor-widget-text-editor\" data-id=\"7e01ecd\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-text-editor.elementor-drop-cap-view-stacked .elementor-drop-cap{background-color:#69727d;color:#fff}.elementor-widget-text-editor.elementor-drop-cap-view-framed .elementor-drop-cap{color:#69727d;border:3px solid;background-color:transparent}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap{margin-top:8px}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap-letter{width:1em;height:1em}.elementor-widget-text-editor .elementor-drop-cap{float:left;text-align:center;line-height:1;font-size:50px}.elementor-widget-text-editor .elementor-drop-cap-letter{display:inline-block}<\/style>\t\t\t\t<p><span style=\"font-family: 'Proxima Nova', sans-serif; font-size: 18px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400;\">Dans\u00a0<\/span><a href=\"https:\/\/datascientest.com\/intelligence-artificielle-quelles-limites\" target=\"_blank\" rel=\"noopener\">un pr\u00e9c\u00e9dent article,<\/a><span style=\"font-family: 'Proxima Nova', sans-serif; font-size: 18px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400;\">\u00a0nous avons discut\u00e9 des limites de l&rsquo;intelligence artificielle en nuan\u00e7ant la vision hollywoodienne d&rsquo;un robot qui d\u00e9passerait l&rsquo;homme et lui dicterait sa conduite.\u00a0\u00a0<\/span><\/p><p><span style=\"font-family: 'Proxima Nova', sans-serif; font-size: 18px; font-style: normal; font-variant-ligatures: normal; font-variant-caps: normal; font-weight: 400;\">N\u00e9anmoins a<\/span><span style=\"font-weight: 400;\">vec l\u2019usage de mod\u00e8les pr\u00e9dictifs g\u00e9n\u00e9ralis\u00e9 au sein de domaines \u00e0 enjeux cruciaux comme le diagnostic m\u00e9dical se pose la question de <\/span><b>la confiance qu\u2019un utilisateur peut accorder \u00e0 une IA<\/b><span style=\"font-weight: 400;\">. En accordant \u00e0 l\u2019intelligence artificielle la possibilit\u00e9 de <\/span><b>guider nos d\u00e9cisions<\/b><span style=\"font-weight: 400;\"> dans un nombre croissant de secteurs comment s\u2019assurer que de tels algorithmes n\u2019engendrent pas des <\/span><b>erreurs trop importantes<\/b><span style=\"font-weight: 400;\"> ?<\/span><\/p><p><b>Et si la solution \u00e9tait dans le probl\u00e8me ? S\u2019il existait une IA capable de justifier les pr\u00e9dictions algorithmiques et de ce fait donner aux mod\u00e8les pr\u00e9dictifs une certaine l\u00e9gitimit\u00e9 et fiabilit\u00e9\u00a0?<\/b><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-a5757df elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"a5757df\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-f52f779\" data-id=\"f52f779\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-974485c elementor-widget elementor-widget-heading\" data-id=\"974485c\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-heading-title{padding:0;margin:0;line-height:1}.elementor-widget-heading .elementor-heading-title[class*=elementor-size-]>a{color:inherit;font-size:inherit;line-height:inherit}.elementor-widget-heading .elementor-heading-title.elementor-size-small{font-size:15px}.elementor-widget-heading .elementor-heading-title.elementor-size-medium{font-size:19px}.elementor-widget-heading .elementor-heading-title.elementor-size-large{font-size:29px}.elementor-widget-heading .elementor-heading-title.elementor-size-xl{font-size:39px}.elementor-widget-heading .elementor-heading-title.elementor-size-xxl{font-size:59px}<\/style><h2 class=\"elementor-heading-title elementor-size-default\">Doit-on se m\u00e9fier des IA ?<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-dfec45d elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"dfec45d\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-7143c44\" data-id=\"7143c44\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-b5211cb elementor-widget elementor-widget-text-editor\" data-id=\"b5211cb\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Pour r\u00e9pondre \u00e0 cette question nous allons nous baser sur un exemple de pr\u00e9diction de deux classifieurs <\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-9aa7955 elementor-widget elementor-widget-image\" data-id=\"9aa7955\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-image{text-align:center}.elementor-widget-image a{display:inline-block}.elementor-widget-image a img[src$=\".svg\"]{width:48px}.elementor-widget-image img{vertical-align:middle;display:inline-block}<\/style>\t\t\t\t\t\t\t\t\t<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/arxiv.org\/pdf\/1602.04938.pdf\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"421\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.09.49-1024x539.png\" class=\"attachment-large size-large wp-image-3400\" alt=\"Algorithme IA\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.09.49-1024x539.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.09.49-300x158.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.09.49-768x404.png 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.09.49.png 1102w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Source: \u201cWhy Should I Trust You?\u201d: Explaining the Predictions of Any Classifier, Tulio Ribeiro, S. Singh, C. Guestrin, ao\u00fbt 2016.<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-568b38c elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"568b38c\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-32b67a1\" data-id=\"32b67a1\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-2a29fd7 elementor-widget elementor-widget-text-editor\" data-id=\"2a29fd7\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Cet exemple pr\u00e9sente les explications des pr\u00e9dictions de deux diff\u00e9rents classifieurs <\/span><span style=\"font-weight: 400;\">tentant de d\u00e9terminer si un document -l\u2019input- traite du \u00ab\u00a0christianisme\u00a0\u00bb ou de \u00ab\u00a0l&rsquo;ath\u00e9isme\u00a0\u00bb.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Dans ces explications, deux \u00e9l\u00e9ments majeurs sont \u00e0 relever\u00a0:\u00a0<\/span><\/p><ul><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Le <\/span><b>graphique<\/b><span style=\"font-weight: 400;\"> en barres repr\u00e9sente l\u2019<\/span><b>ordre d\u2019importance des mots <\/b><span style=\"font-weight: 400;\">conduisant \u00e0 la pr\u00e9diction\u00a0<\/span><\/li><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">La <\/span><b>couleur <\/b><span style=\"font-weight: 400;\">indique \u00e0 quelle classe le mot contribue : vert pour \u00ab\u00a0christianisme\u00a0\u00bb et magenta pour \u00ab\u00a0ath\u00e9isme\u00a0\u00bb<\/span><\/li><\/ul><p><span style=\"font-weight: 400;\">Cette exp\u00e9rience montre que dans ce cas, <\/span><b>les pr\u00e9dictions sont faites pour des raisons assez arbitraires\u00a0<\/b><span style=\"font-weight: 400;\">: pour l\u2019algorithme 2, les mots \u00ab\u00a0Post\u00a0\u00bb, \u00ab\u00a0Host\u00a0\u00bb et \u00ab\u00a0Re\u00a0\u00bb, contrairement \u00e0 ce que le graphique en barre pr\u00e9sente, n&rsquo;ont aucun lien avec l&rsquo;ath\u00e9isme (ni le christianisme d\u2019ailleurs), tandis que pour l\u2019algorithme 1, les termes \u201cGOD\u201c ou \u201cKoresh\u201c ne devraient pas conduire \u00e0 l\u2019ath\u00e9isme mais plut\u00f4t au christianisme.<\/span><\/p><blockquote><p><span style=\"font-weight: 400;\">De ce fait, comme nous montre cet exemple, avoir une confiance aveugle envers une IA serait une erreur \u00e0 ne pas commettre. <\/span><span style=\"font-weight: 400;\">Comment savoir si une IA prend une d\u00e9cision pour une raison sens\u00e9e\u00a0?<\/span><\/p><\/blockquote><p><span style=\"font-weight: 400;\">Nous allons vous pr\u00e9senter maintenant <\/span><b>LIME, une librairie disponible sur Python<\/b><span style=\"font-weight: 400;\">\u00a0permettant de <\/span><b>transformer tout mod\u00e8le complexe<\/b><span style=\"font-weight: 400;\"> et donc <\/span><b>difficilement interpr\u00e9table<\/b> <b>en un mod\u00e8le lin\u00e9aire simple <\/b><span style=\"font-weight: 400;\">dont l\u2019avantage \u00e9vident est qu\u2019il sera d\u00e9sormais plus facile d\u2019expliquer la pr\u00e9diction.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-2d7b82b elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"2d7b82b\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-9671b56\" data-id=\"9671b56\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-d9da9f1 elementor-widget elementor-widget-heading\" data-id=\"d9da9f1\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">L\u2019approche LIME : une solution efficace ?<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-19cc916 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"19cc916\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-a6e3da7\" data-id=\"a6e3da7\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-6bff6fc elementor-widget elementor-widget-text-editor\" data-id=\"6bff6fc\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">LIME (Locally Interpretable Model-agnostic Explanation) est une IA qui permet de <\/span><b>comprendre la logique sous-jacente des pr\u00e9dictions algorithmiques<\/b><span style=\"font-weight: 400;\"> afin d&rsquo;accompagner l&rsquo;humain dans sa d\u00e9cision \u00e0 croire ou non en ces pr\u00e9visions.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Si certains mod\u00e8les de par leur simplicit\u00e9 (r\u00e9gression lin\u00e9aire, arbres de d\u00e9cision) permettent d&rsquo;expliquer leur pr\u00e9diction, d\u2019autres<\/span><b> algorithmes de machine learning<\/b><span style=\"font-weight: 400;\"> tels que ceux reposant sur l\u2019<\/span><b>architecture des r\u00e9seaux de neurones<\/b><span style=\"font-weight: 400;\"> disposent g\u00e9n\u00e9ralement d\u2019une fonction de pr\u00e9diction d\u2019une <\/span><b>complexit\u00e9 trop importante<\/b><span style=\"font-weight: 400;\"> et restent donc inintelligibles pour l\u2019Homme. On parle de <\/span><b>mod\u00e8les difficilement interpr\u00e9tables<\/b><span style=\"font-weight: 400;\">.<\/span><\/p><p><span style=\"font-weight: 400;\">L\u2019objectif de LIME est de <\/span><b>convertir les pr\u00e9dictions en un mod\u00e8le math\u00e9matique lin\u00e9aire<\/b><span style=\"font-weight: 400;\"> interpr\u00e9table. Autrement dit, le but de LIME est de <\/span><b>trouver la meilleure version simplifi\u00e9e d\u2019un mod\u00e8le complexe<\/b><span style=\"font-weight: 400;\"> autour d&rsquo;une seule observation \u00e9tudi\u00e9e. La figure ci-dessous pr\u00e9sente la nouvelle classification lin\u00e9aire fa\u00eetes suite \u00e0 l\u2019utilisation de LIME\u00a0:<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-6a0eac3 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"6a0eac3\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-f163921\" data-id=\"f163921\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-74a2efc elementor-widget elementor-widget-image\" data-id=\"74a2efc\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"492\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.11.png\" class=\"attachment-large size-large wp-image-3401\" alt=\"Algorithme IA LIME\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.11.png 1002w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.11-300x184.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.11-768x472.png 768w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Source: \u201cWhy Should I Trust You?\u201d: Explaining the Predictions of Any Classifier<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-163781f elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"163781f\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-559b925\" data-id=\"559b925\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-3e39141 elementor-widget elementor-widget-text-editor\" data-id=\"3e39141\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Initialement, le mod\u00e8le fait une pr\u00e9diction binaire complexe (zone rose ou bleue)\u00a0:\u00a0<\/span><\/p><ul><li><span style=\"font-weight: 400;\">La <\/span><b>croix rouge<\/b><span style=\"font-weight: 400;\"> en gras correspond \u00e0 la <\/span><b>pr\u00e9diction<\/b><span style=\"font-weight: 400;\">.<\/span><\/li><li><span style=\"font-weight: 400;\">Les<\/span><b> axes<\/b><span style=\"font-weight: 400;\"> sont les <\/span><b>variables explicatives<\/b><span style=\"font-weight: 400;\"> (features).<\/span><\/li><li><span style=\"font-weight: 400;\">Les <\/span><b>ronds bleus et croix rouges<\/b><span style=\"font-weight: 400;\"> sont les <\/span><b>pr\u00e9dictions obtenues apr\u00e8s perturbations des features<\/b><span style=\"font-weight: 400;\">.<\/span><\/li><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Tout point situ\u00e9 \u00e0 la droite de la pr\u00e9diction initiale a vu sa feature correspondant \u00e0 l\u2019axe des abscisses modifi\u00e9e et celle des ordonn\u00e9es inchang\u00e9e.\u00a0<\/span><\/li><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Plus un point est proche de la pr\u00e9diction initiale (croix rouge en gras), plus il poss\u00e8de une grande taille.<\/span><\/li><\/ul><blockquote><span style=\"font-weight: 400;\">Le graphique ci-dessus met ainsi en lumi\u00e8re le r\u00f4le de LIME \u00e0 savoir nous ramener \u00e0 un mod\u00e8le lin\u00e9aire simple dont la fronti\u00e8re est repr\u00e9sent\u00e9e par la droite en pointill\u00e9\u00a0: contrairement au mod\u00e8le binaire initial complexe (r\u00e9gion rose ou bleue), d\u00e9sormais les \u201cnouvelles\u201c zones rouges et bleues seront celles situ\u00e9es respectivement \u00e0 la gauche et \u00e0 la droite des pointill\u00e9s.<\/span><\/blockquote><p><span style=\"font-weight: 400;\">Par ailleurs, la droite repr\u00e9sent\u00e9e en pointill\u00e9 d\u00e9termin\u00e9e par LIME r\u00e9v\u00e8le \u00e9galement que<\/span><b> la variable sur l\u2019axe des abscisses est plus importante que celle de l\u2019axe des ordonn\u00e9es<\/b><span style=\"font-weight: 400;\"> :\u00a0<\/span><\/p><ul><li><span style=\"font-weight: 400;\">Un d\u00e9placement sur l\u2019axe des ordonn\u00e9es a peu d\u2019impact sur la pr\u00e9diction (le point reste dans la r\u00e9gion rouge).\u00a0<\/span><\/li><li><span style=\"font-weight: 400;\">Un d\u00e9placement sur l\u2019axe des abscisses modifie la pr\u00e9diction (le point passe de la r\u00e9gion rouge \u00e0 bleue).<\/span><\/li><li style=\"list-style-type: none;\">\u00a0<\/li><\/ul>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7b99473 elementor-widget elementor-widget-heading\" data-id=\"7b99473\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h2 class=\"elementor-heading-title elementor-size-default\">Exemple d\u2019utilisation pratique de LIME :\n<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-f2dd22f elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"f2dd22f\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-e1b3ba6\" data-id=\"e1b3ba6\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-43abd33 elementor-widget elementor-widget-image\" data-id=\"43abd33\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"419\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.20.png\" class=\"attachment-large size-large wp-image-3402\" alt=\"Algorithme IA\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.20.png 894w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.20-300x157.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.20-768x402.png 768w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-9c81141 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"9c81141\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-4630229\" data-id=\"4630229\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-b3932f6 elementor-widget elementor-widget-text-editor\" data-id=\"b3932f6\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Dans cet exemple, <\/span><b>LIME<\/b><span style=\"font-weight: 400;\">, en transformant un mod\u00e8le complexe quelconque en un mod\u00e8le lin\u00e9aire interpr\u00e9table, <\/span><b>justifie la pr\u00e9diction<\/b><span style=\"font-weight: 400;\"> d\u2019un <\/span><b>algorithme de reconnaissance d\u2019image <\/b><span style=\"font-weight: 400;\">en<\/span><b> d\u00e9tectant pr\u00e9cis\u00e9ment quelle portion de l\u2019image fournie en entr\u00e9e a permis de reconna\u00eetre la pr\u00e9sence d\u2019une grenouille<\/b><span style=\"font-weight: 400;\">.<\/span><\/p><p><span style=\"font-weight: 400;\">A travers cet exemple, on comprend que LIME a \u00e9t\u00e9 con\u00e7u dans le but de permettre aux utilisateurs d\u2019<\/span><b>avoir confiance aux pr\u00e9dictions des IA<\/b><span style=\"font-weight: 400;\">. On peut donc penser que dans un futur proche, les IA seront davantage soumises \u00e0 la contrainte de totale transparence envers les utilisateurs, surtout depuis l\u2019arriv\u00e9e du<\/span><a href=\"https:\/\/www.cnil.fr\/fr\/comprendre-le-rgpd\" target=\"_blank\" rel=\"noopener\">\u00a0RGPD<\/a><span style=\"font-weight: 400;\">\u00a0(R\u00e8glement G\u00e9n\u00e9ral sur la Protection des Donn\u00e9es) en mai 2018. Il sera alors n\u00e9cessaire de<\/span><b> cr\u00e9er de nouveaux algorithmes de plus en plus performants permettant de justifier les d\u00e9cisions et pr\u00e9dictions de mod\u00e8les complexes<\/b><span style=\"font-weight: 400;\"> et par cons\u00e9quent, difficilement interpr\u00e9tables.<\/span><\/p><p>\u00a0<\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-1eb55bc elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"1eb55bc\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-c52730d\" data-id=\"c52730d\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-6d5edab elementor-headline--style-highlight elementor-widget elementor-widget-animated-headline\" data-id=\"6d5edab\" data-element_type=\"widget\" data-settings=\"{&quot;marker&quot;:&quot;double_underline&quot;,&quot;highlighted_text&quot;:&quot;colossal &quot;,&quot;headline_style&quot;:&quot;highlight&quot;,&quot;loop&quot;:&quot;yes&quot;,&quot;highlight_animation_duration&quot;:1200,&quot;highlight_iteration_delay&quot;:8000}\" data-widget_type=\"animated-headline.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<link rel=\"stylesheet\" href=\"https:\/\/datascientest.com\/wp-content\/plugins\/elementor-pro\/assets\/css\/widget-animated-headline.min.css\">\t\t<h3 class=\"elementor-headline\">\n\t\t\t\t\t<span class=\"elementor-headline-plain-text elementor-headline-text-wrapper\">Un travail <\/span>\n\t\t\t\t<span class=\"elementor-headline-dynamic-wrapper elementor-headline-text-wrapper\">\n\t\t\t\t\t<span class=\"elementor-headline-dynamic-text elementor-headline-text-active\">colossal <\/span>\n\t\t\t\t<\/span>\n\t\t\t\t\t<span class=\"elementor-headline-plain-text elementor-headline-text-wrapper\">en perspective\u00a0!<\/span>\n\t\t\t\t\t<\/h3>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-e3277bb elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"e3277bb\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-7e9df3d\" data-id=\"7e9df3d\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-dcf66ce elementor-widget elementor-widget-text-editor\" data-id=\"dcf66ce\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p>Cet article vous a plu ? Vous souhaitez approfondir vos connaissances en IA?&nbsp;<\/p>\n<p><a href=\"https:\/\/datascientest.com\/#dateformations\" target=\"_blank\" style=\"font-size: 18px; background-color: rgb(255, 255, 255);\">D\u00e9marrez prochainement l&rsquo;une de formations<\/a>&nbsp;!&nbsp;<\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>Dans&nbsp;un pr\u00e9c\u00e9dent article,&nbsp;nous avons discut\u00e9 des limites de l&rsquo;intelligence artificielle en nuan\u00e7ant la vision hollywoodienne d&rsquo;un robot qui d\u00e9passerait l&rsquo;homme et lui dicterait sa conduite.&nbsp;&nbsp; N\u00e9anmoins avec l\u2019usage de mod\u00e8les pr\u00e9dictifs g\u00e9n\u00e9ralis\u00e9 au sein de domaines \u00e0 enjeux cruciaux comme le diagnostic m\u00e9dical se pose la question de la confiance qu\u2019un utilisateur peut accorder \u00e0 [&hellip;]<\/p>\n","protected":false},"author":77,"featured_media":3401,"comment_status":"open","ping_status":"open","sticky":false,"template":"elementor_theme","format":"standard","meta":{"content-type":"","inline_featured_image":false},"categories":[2497],"tags":[156],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v21.7 (Yoast SEO v21.7) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Peut-on avoir confiance aux pr\u00e9dictions d\u2019une IA : L\u2019approche LIME<\/title>\n<meta name=\"description\" content=\"Comment faire confiance aux predictions d&#039;une IA ? D\u00e9couvrez une approche originale dans cet article, une IA pour solutionner le probl\u00e8me!\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Peut-on avoir confiance aux pr\u00e9dictions d\u2019une IA : L\u2019approche LIME\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2020-06-08T16:14:09+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-11-09T14:14:17+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.11.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1002\" \/>\n\t<meta property=\"og:image:height\" content=\"616\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Raphael Kassel\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"Raphael Kassel\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"5 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime\",\"url\":\"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime\",\"name\":\"Peut-on avoir confiance aux pr\u00e9dictions d\u2019une IA : L\u2019approche LIME\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2020-06-08T16:14:09+00:00\",\"dateModified\":\"2023-11-09T14:14:17+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1\"},\"description\":\"Comment faire confiance aux predictions d'une IA ? D\u00e9couvrez une approche originale dans cet article, une IA pour solutionner le probl\u00e8me!\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime\"]}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/datascientest.com\/#website\",\"url\":\"https:\/\/datascientest.com\/\",\"name\":\"Formation Data Science | DataScientest.com\",\"description\":\"Leader de la formation en Data Science\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/datascientest.com\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1\",\"name\":\"Raphael Kassel\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg\",\"contentUrl\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg\",\"caption\":\"Raphael Kassel\"},\"description\":\"Raphael Kassel est le Head of Data Programs chez DataScientest. Il supervise une \u00e9quipe Data, impl\u00e9mente de nouveaux parcours comme \\\"Chef de projet IA\\\" et \\\"Data Product Manager\\\", et contribue \u00e0 la modernisation des op\u00e9rations du groupe. En collaboration \u00e9troite avec l'\u00e9quipe de d\u00e9veloppement, il adapte l'accompagnement acad\u00e9mique aux \u00e9volutions de l'\u00e9quipe data et \u00e0 l'internationalisation.\",\"sameAs\":[\"https:\/\/www.linkedin.com\/in\/raphaelkassel\/\",\"Data Science\"],\"url\":\"https:\/\/datascientest.com\/author\/raphaelkassel\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Peut-on avoir confiance aux pr\u00e9dictions d\u2019une IA : L\u2019approche LIME","description":"Comment faire confiance aux predictions d'une IA ? D\u00e9couvrez une approche originale dans cet article, une IA pour solutionner le probl\u00e8me!","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime","og_locale":"fr_FR","og_type":"article","og_title":"Peut-on avoir confiance aux pr\u00e9dictions d\u2019une IA : L\u2019approche LIME","og_url":"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2020-06-08T16:14:09+00:00","article_modified_time":"2023-11-09T14:14:17+00:00","og_image":[{"width":1002,"height":616,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Capture-d\u2019e\u0301cran-2020-06-08-a\u0300-18.11.11.png","type":"image\/png"}],"author":"Raphael Kassel","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"Raphael Kassel","Dur\u00e9e de lecture estim\u00e9e":"5 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime","url":"https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime","name":"Peut-on avoir confiance aux pr\u00e9dictions d\u2019une IA : L\u2019approche LIME","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2020-06-08T16:14:09+00:00","dateModified":"2023-11-09T14:14:17+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1"},"description":"Comment faire confiance aux predictions d'une IA ? D\u00e9couvrez une approche originale dans cet article, une IA pour solutionner le probl\u00e8me!","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/peut-on-avoir-confiance-aux-predictions-dune-ia-lapproche-lime"]}]},{"@type":"WebSite","@id":"https:\/\/datascientest.com\/#website","url":"https:\/\/datascientest.com\/","name":"Formation Data Science | DataScientest.com","description":"Leader de la formation en Data Science","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/datascientest.com\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"fr-FR"},{"@type":"Person","@id":"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1","name":"Raphael Kassel","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/datascientest.com\/#\/schema\/person\/image\/","url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg","contentUrl":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg","caption":"Raphael Kassel"},"description":"Raphael Kassel est le Head of Data Programs chez DataScientest. Il supervise une \u00e9quipe Data, impl\u00e9mente de nouveaux parcours comme \"Chef de projet IA\" et \"Data Product Manager\", et contribue \u00e0 la modernisation des op\u00e9rations du groupe. En collaboration \u00e9troite avec l'\u00e9quipe de d\u00e9veloppement, il adapte l'accompagnement acad\u00e9mique aux \u00e9volutions de l'\u00e9quipe data et \u00e0 l'internationalisation.","sameAs":["https:\/\/www.linkedin.com\/in\/raphaelkassel\/","Data Science"],"url":"https:\/\/datascientest.com\/author\/raphaelkassel"}]}},"_links":{"self":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/3384"}],"collection":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/users\/77"}],"replies":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/comments?post=3384"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/3384\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/3401"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=3384"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=3384"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=3384"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}