{"id":6629,"date":"2020-07-31T09:17:11","date_gmt":"2020-07-31T09:17:11","guid":{"rendered":"https:\/\/datascientest.com\/?p=6629"},"modified":"2023-11-09T14:12:48","modified_gmt":"2023-11-09T14:12:48","slug":"reinforcement-learning","status":"publish","type":"post","link":"https:\/\/datascientest.com\/reinforcement-learning","title":{"rendered":"Reinforcement Learning : D\u00e9finition et application"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"6629\" class=\"elementor elementor-6629\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-a87b5d8 e-flex e-con-boxed e-con\" data-id=\"a87b5d8\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-b077944 elementor-widget elementor-widget-heading\" data-id=\"b077944\" 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\"><p>Le Reinforcement Learning, ou apprentissage par renforcement en fran\u00e7ais, suscite depuis quelques ann\u00e9es un tr\u00e8s grand int\u00e9r\u00eat. On se souvient encore de ce moment historique o\u00f9 AlphaGo, une machine d\u2019Intelligence Artificielle, a r\u00e9ussi \u00e0 vaincre un champion humain incontest\u00e9 !<\/p> \n<p>Depuis, le Reinforcement Learning a r\u00e9alis\u00e9 d\u2019importants progr\u00e8s, surpassant les m\u00e9thodes bien \u00e9tablies. N\u00e9anmoins, l\u2019utilisation du Reinforcement Learning dans le monde professionnel est loin d\u2019\u00eatre g\u00e9n\u00e9ralis\u00e9e. Malgr\u00e9 des performances prometteuses, tr\u00e8s peu d\u2019entreprises ont r\u00e9ussi \u00e0 identifier des cas d\u2019applications concrets et l\u2019impl\u00e9menter pour am\u00e9liorer leurs process.<\/p><\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-e0b9626 elementor-widget elementor-widget-text-editor\" data-id=\"e0b9626\" 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-weight: 400;\">Sur quels principes repose le Reinforcement Learning\u00a0? Comment se distingue-t-il des autres modes d\u2019apprentissage\u00a0? Quels sont les domaines d\u2019application auxquels il s\u2019applique\u00a0? Cet article vous dit tout sur le potentiel de l&rsquo;apprentissage par renforcement.<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-a72a204 e-flex e-con-boxed e-con\" data-id=\"a72a204\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-781691d elementor-widget elementor-widget-heading\" data-id=\"781691d\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Qu'est ce que le Reinforcement Learning ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2041319 elementor-widget elementor-widget-text-editor\" data-id=\"2041319\" 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><b>Le Reinforcement Learning d\u00e9signe l\u2019ensemble des m\u00e9thodes qui permettent \u00e0 un agent d\u2019apprendre \u00e0 choisir quelle action prendre, et ceci de mani\u00e8re autonome<\/b><span style=\"font-weight: 400;\">.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Plong\u00e9 dans un environnement donn\u00e9, il apprend en recevant des r\u00e9compenses ou des p\u00e9nalit\u00e9s en fonction de ses actions. Au travers de son exp\u00e9rience, l\u2019agent cherche \u00e0 trouver la <\/span><b>strat\u00e9gie d\u00e9cisionnelle optimale<\/b><span style=\"font-weight: 400;\"> qui puisse lui permettre de <\/span><b>maximiser les r\u00e9compenses <\/b><span style=\"font-weight: 400;\">accumul\u00e9es au cours du temps.<\/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-6abe65e elementor-widget elementor-widget-image\" data-id=\"6abe65e\" 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:\/\/en.wikipedia.org\/wiki\/Reinforcement_learning\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"996\" height=\"778\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Plan-de-travail-2.png\" class=\"attachment-full size-full wp-image-219664\" alt=\"Sch\u00e9ma d\u2019apprentissage par renforcement, adapt\u00e9 de [5]\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Plan-de-travail-2.png 996w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Plan-de-travail-2-300x234.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Plan-de-travail-2-768x600.png 768w\" sizes=\"(max-width: 996px) 100vw, 996px\" \/>\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\">Sch\u00e9ma d\u2019apprentissage par renforcement, adapt\u00e9 de [5]<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-03352ca e-flex e-con-boxed e-con\" data-id=\"03352ca\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-e63fe1e elementor-widget elementor-widget-heading\" data-id=\"e63fe1e\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Comment formuler un probl\u00e8me de Reinforcement learning ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-3000f27 elementor-widget elementor-widget-text-editor\" data-id=\"3000f27\" 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>L&rsquo;apprentissage par renforcement\u00a0n\u00e9cessite d\u2019introduire un certain nombre de concepts et de m\u00e9triques, dont les principaux sont les suivants\u00a0:\u00a0\u00a0<\/p><ul><li><strong>Agent\u00a0<\/strong>: syst\u00e8me ou robot qui interagit et agit dans l\u2019environnement\u00a0;<\/li><li><strong>Action a<\/strong>\u00a0: une action parmi l\u2019ensemble des actions possible\u00a0;<\/li><li><strong>\u00c9tat s<\/strong> : situation particuli\u00e8re dans laquelle l\u2019agent se trouve ;<\/li><li><strong>Politique \u03c0<\/strong> : strat\u00e9gie qui d\u00e9finit le comportement l\u2019agent<\/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-a181952 elementor-widget elementor-widget-text-editor\" data-id=\"a181952\" 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>Dans le cas d\u2019une politique d\u00e9terministe, l\u2019action a est d\u00e9finie par\u00a0: a= \u03c0(a|s)<\/p><p>Dans le cas d\u2019une politique stochastique, la probabilit\u00e9 de l\u2019action a est d\u00e9finie par\u00a0:p(a|s)= \u03c0(a|s)<\/p><ul><li><strong>R\u00e9compense r<\/strong>(s,a)\u00a0: Gain positif ou n\u00e9gatif collect\u00e9 en effectuant l\u2019action a \u00e0 l\u2019\u00e9tat s. L\u2019objectif est de maximiser les b\u00e9n\u00e9fices totaux d&rsquo;une politique.\u00a0<\/li><li><strong>\u00c9pisode\u00a0<\/strong>: il peut \u00eatre d\u00e9fini comme la suite des actions men\u00e9es jusqu\u2019\u00e0 l\u2019\u00e9tat final ou une dur\u00e9e d\u2019action pr\u00e9d\u00e9finie.\u00a0<\/li><li><strong>Fonction de valeur V(s)<\/strong> : La fonction de valeur d&rsquo;un \u00e9tat s est le montant total des r\u00e9compenses qu&rsquo;un agent s\u2019attend \u00e0 pouvoir collecter de cet \u00e9tat jusqu&rsquo;\u00e0 la fin de l&rsquo;\u00e9pisode.<\/li><li><strong>Fonction action-valeur Q(s, a)\u00a0<\/strong>: La fonction de valeur d&rsquo;action a \u00e0 l\u2019\u00e9tat s est le montant total des r\u00e9compenses attendues en prenant l&rsquo;action a \u00e0 l\u2019\u00e9tat s jusqu&rsquo;\u00e0 la fin de l&rsquo;\u00e9pisode.<\/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-0452249 elementor-widget elementor-widget-text-editor\" data-id=\"0452249\" 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;\">Prenons l\u2019exemple du jeu de <\/span><b>Pac-Man<\/b><span style=\"font-weight: 400;\">. Le but de l&rsquo;agent (Pac-Man) est de manger la nourriture dans la grille tout en \u00e9vitant les fant\u00f4mes sur son chemin. Le <\/span><b>monde de la grille<\/b><span style=\"font-weight: 400;\"> est l&rsquo;<\/span><b>environnement <\/b><span style=\"font-weight: 400;\">interactif de l&rsquo;agent. Pac-Man re\u00e7oit une r\u00e9compense pour avoir mang\u00e9 de la nourriture et une punition s&rsquo;il est tu\u00e9 par le fant\u00f4me (il perd le jeu). Les <\/span><b>\u00e9tats <\/b><span style=\"font-weight: 400;\">correspondent \u00e0 l&rsquo;<\/span><b>emplacement <\/b><span style=\"font-weight: 400;\">de Pac-Man dans le monde de la grille et la r\u00e9compense totale cumul\u00e9e est la victoire de Pac-Man dans le jeu.<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-cb97da4 e-flex e-con-boxed e-con\" data-id=\"cb97da4\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-9312c0e elementor-widget elementor-widget-heading\" data-id=\"9312c0e\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">En quoi le Reinforcement Learning est diff\u00e9rent des autres m\u00e9thodes d\u2019apprentissage ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-962cb75 elementor-widget elementor-widget-text-editor\" data-id=\"962cb75\" 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><b>De mani\u00e8re g\u00e9n\u00e9rale, l\u2019apprentissage automatique correspond \u00e0 l\u2019apprentissage r\u00e9alis\u00e9 par des algorithmes \u00e0 partir de donn\u00e9es en vue de r\u00e9aliser des pr\u00e9dictions ou de prendre des d\u00e9cisions.\u00a0<\/b><\/p><p><span style=\"font-weight: 400;\">En ce sens, l\u2019<\/span><b>apprentissage par renforcement<\/b><span style=\"font-weight: 400;\"> repr\u00e9sente l\u2019<\/span><b>un des trois paradigmes d\u2019apprentissage automatique <\/b><span style=\"font-weight: 400;\">aux c\u00f4t\u00e9s de l\u2019<\/span><b>apprentissage supervis\u00e9 <\/b><span style=\"font-weight: 400;\">et l\u2019<\/span><b>apprentissage non-supervis\u00e9<\/b><span style=\"font-weight: 400;\">.<\/span><\/p><ul><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Dans le cadre de l&rsquo;<\/span><b>apprentissage supervis\u00e9,<\/b><span style=\"font-weight: 400;\"> les <\/span><b>donn\u00e9es sont labellis\u00e9es<\/b><span style=\"font-weight: 400;\">\u00a0: chaque observation en entr\u00e9e est associ\u00e9e \u00e0 une variable de sortie. <\/span><b>L\u2019objectif de l\u2019algorithme est donc d\u2019apprendre \u00e0 associer observations et labels \u00e0 partir de ces donn\u00e9es.<\/b><\/li><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">En ce qui concerne l&rsquo;<\/span><b>apprentissage non supervis\u00e9<\/b><span style=\"font-weight: 400;\">, les <\/span><b>donn\u00e9es ne sont pas labellis\u00e9es<\/b><span style=\"font-weight: 400;\">. L\u2019algorithme a alors pour but d\u2019<\/span><b>identifier des caract\u00e9ristiques communes<\/b><span style=\"font-weight: 400;\"> au sein des donn\u00e9es et donc de <\/span><b>d\u00e9couvrir des structures<\/b> <b>sous-jacentes<\/b><span style=\"font-weight: 400;\">.<\/span><\/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-9678428 elementor-widget elementor-widget-text-editor\" data-id=\"9678428\" 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;\">En revanche, dans le cas de l\u2019<\/span><b>apprentissage par renforcement<\/b><span style=\"font-weight: 400;\">, les <\/span><b>observations re\u00e7ues par l\u2019algorithme sont associ\u00e9es \u00e0 des retours \u00e9valuatifs fournis par l\u2019environnement<\/b><span style=\"font-weight: 400;\"> et non pas des labels.<\/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-2bc992e eael-table-align-center eael-dt-th-align-left elementor-widget elementor-widget-eael-data-table\" data-id=\"2bc992e\" data-element_type=\"widget\" data-widget_type=\"eael-data-table.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"eael-data-table-wrap\" data-table_id=\"2bc992e\" id=\"eael-data-table-wrapper-2bc992e\" data-custom_responsive=\"false\">\n\t\t\t<table class=\"tablesorter eael-data-table center\" id=\"eael-data-table-2bc992e\">\n\t\t\t    <thead>\n\t\t\t        <tr class=\"table-header\">\n\t\t\t\t\t\t\t\t\t            <th class=\"\" id=\"\" colspan=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t<span class=\"data-table-header-text\"><\/span><\/th>\n\t\t\t        \t\t\t\t            <th class=\"\" id=\"\" colspan=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t<span class=\"data-table-header-text\">Apprentissage supervis\u00e9<\/span><\/th>\n\t\t\t        \t\t\t\t            <th class=\"\" id=\"\" colspan=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t<span class=\"data-table-header-text\">Apprentissage non-supervis\u00e9<\/span><\/th>\n\t\t\t        \t\t\t\t            <th class=\"\" id=\"\" colspan=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t<span class=\"data-table-header-text\">Apprentissage par renforcement<\/span><\/th>\n\t\t\t        \t\t\t\t        <\/tr>\n\t\t\t    <\/thead>\n\t\t\t  \t<tbody>\n\t\t\t\t\t\t\t\t\t\t\t<tr>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\"><i><b>D\u00e9finition<\/b><\/i><\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">L'algorithme apprend \u00e0 partir de donn\u00e9es labellis\u00e9es<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">L'algorithme est entrain\u00e9 \u00e0 partir de donn\u00e9es non labellis\u00e9es sans indications particuli\u00e8res<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">L'algorithme interagit avec son environnement en r\u00e9alisant des actions et en apprenant de ses erreurs et succ\u00e8s<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/tr>\n\t\t\t        \t\t\t\t\t\t<tr>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\"><i><b>Types de probl\u00e8mes<\/b><\/i><\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">R\u00e9gression et classification<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">Association et clustering<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">Bas\u00e9 sur un syst\u00e8me de r\u00e9compense<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/tr>\n\t\t\t        \t\t\t\t\t\t<tr>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\"><i><b>Type de donn\u00e9es<\/b><\/i><\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">Donn\u00e9es labellis\u00e9es<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">Donn\u00e9es non labellis\u00e9es<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">Pas de donn\u00e9es fournies au pr\u00e9alable<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/tr>\n\t\t\t        \t\t\t\t\t\t<tr>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\"><i><b>Approche<\/b><\/i><\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">Etudie les relations sous-jacentes qui lient les donn\u00e9es en entr\u00e9e aux labels<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">D\u00e9couvre les motifs communs au sein des donn\u00e9es d'entr\u00e9e<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t   \t\t\t\t\t\t\t\t\t\t\t<td colspan=\"\" rowspan=\"\" class=\"\" id=\"\">\n\t\t\t\t\t\t\t\t\t\t\t\t<div class=\"td-content-wrapper\"><div class=\"td-content\">Apprend une strat\u00e9gie de comportement en fonction d'exp\u00e9riences pass\u00e9es et des r\u00e9compenses per\u00e7ues<\/div><\/div>\n\t\t\t\t\t\t\t\t\t\t\t<\/td>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/tr>\n\t\t\t        \t\t\t    <\/tbody>\n\t\t\t<\/table>\n\t\t<\/div>\n\t  \t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-fa9ce20 elementor-widget elementor-widget-text-editor\" data-id=\"fa9ce20\" 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 style=\"text-align: center;\"><em>Comparaison entre les diff\u00e9rents types d\u2019apprentissage, adapt\u00e9 de [4]<\/em><\/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-0a2f414 elementor-hidden-desktop elementor-hidden-tablet elementor-hidden-mobile elementor-widget elementor-widget-image\" data-id=\"0a2f414\" 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=\"1024\" height=\"768\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/meth-dapprentissage-RL.png\" class=\"attachment-full size-full wp-image-6635\" alt=\"Apprentissage par renforcement\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/meth-dapprentissage-RL.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/meth-dapprentissage-RL-300x225.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/meth-dapprentissage-RL-768x576.png 768w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Comparaison entre les diff\u00e9rents types d\u2019apprentissage, adapt\u00e9 de [4]<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-40a50d7 e-flex e-con-boxed e-con\" data-id=\"40a50d7\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-8bd636b elementor-widget elementor-widget-heading\" data-id=\"8bd636b\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Quels les principaux challenges du Reinforcement Learning ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-e45f114 elementor-widget elementor-widget-heading\" data-id=\"e45f114\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-default\">Le dilemme entre l\u2019exploration et l\u2019exploitation<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1725202 elementor-widget elementor-widget-text-editor\" data-id=\"1725202\" 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;\">Lorsque l\u2019agent est plong\u00e9 dans un environnement inconnu, il doit concilier entre\u00a0:<\/span><\/p><ol><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">La n\u00e9cessit\u00e9 d\u2019explorer afin d\u2019acqu\u00e9rir une meilleure connaissance de son environnement\u00a0<\/span><\/li><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">L\u2019exploitation de ses connaissances pour maximiser ses gains<\/span><\/li><\/ol>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-76319a7 elementor-widget elementor-widget-text-editor\" data-id=\"76319a7\" 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 \u00e9quilibre entre exploitation et exploration repr\u00e9sente un enjeu essentiel dans la <\/span><b>phase d\u2019apprentissage<\/b><span style=\"font-weight: 400;\">. \u00c0 titre d\u2019exemple, dans le cas de PacMan, l&rsquo;agent est confront\u00e9 au dilemme d&rsquo;explorer de nouveaux \u00e9tats en parcourant la grille ou de maximiser les r\u00e9compenses accumul\u00e9es en limitant ses d\u00e9placements pour \u00e9viter les fant\u00f4mes.<\/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-21e6d3b elementor-widget elementor-widget-heading\" data-id=\"21e6d3b\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-default\">Choisir l'algorithme appropri\u00e9 et en r\u00e9gler les param\u00e8tres<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-0ac137d elementor-widget elementor-widget-text-editor\" data-id=\"0ac137d\" 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;\">Le <\/span><b>choix de l&rsquo;algorithme<\/b><span style=\"font-weight: 400;\"> le plus appropri\u00e9 repr\u00e9sente un d\u00e9fi important \u00e9tant donn\u00e9 la grande vari\u00e9t\u00e9 d&rsquo;approches existantes. Celles-ci reposent sur des <\/span><b>m\u00e9thodes diff\u00e9rentes<\/b><span style=\"font-weight: 400;\"> et utilisent de nombreux <\/span><b>hyper-param\u00e8tres<\/b><span style=\"font-weight: 400;\">. Dans ce contexte, il est essentiel de d\u00e9finir les bonnes mesures pour \u00e9valuer ses performances et de choisir avec soin ses hyper-param\u00e8tres.<\/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-4180ce3 elementor-widget elementor-widget-heading\" data-id=\"4180ce3\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-default\">Faire face aux probl\u00e9matiques de g\u00e9n\u00e9ralisation<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-52e7ff6 elementor-widget elementor-widget-text-editor\" data-id=\"52e7ff6\" 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;\">Comme dans tout autre domaine de l&rsquo;<\/span><b>apprentissage machine<\/b><span style=\"font-weight: 400;\">, les algorithmes d&rsquo;apprentissage par renforcement sont confront\u00e9s \u00e0 des probl\u00e8mes de g\u00e9n\u00e9ralisation. Plus pr\u00e9cis\u00e9ment, le concept de g\u00e9n\u00e9ralisation renvoie \u00e0 deux crit\u00e8res :\u00a0<\/span><\/p><ul><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Atteindre des <\/span><b>performances \u00e9lev\u00e9es<\/b><span style=\"font-weight: 400;\"> dans un environnement o\u00f9 les <\/span><b>donn\u00e9es recueillies <\/b><span style=\"font-weight: 400;\">sont <\/span><b>limit\u00e9es<\/b><\/li><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Obtenir des performances \u00e9lev\u00e9es dans un environnement aux caract\u00e9ristiques similaires<\/span><\/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-e15024f elementor-widget elementor-widget-text-editor\" data-id=\"e15024f\" 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 mesurer la performance de la g\u00e9n\u00e9ralisation d&rsquo;une m\u00e9thode dans le premier cas, le comportement de l&rsquo;agent est observ\u00e9 dans un environnement de test identique \u00e0 celui sur lequel il a \u00e9t\u00e9 form\u00e9. En revanche, dans le second cas, le <\/span><b>comportement de l&rsquo;agent<\/b><span style=\"font-weight: 400;\"> est observ\u00e9 dans un <\/span><b>environnement de test qui a des caract\u00e9ristiques communes <\/b><span style=\"font-weight: 400;\">avec l&rsquo;environnement de formation mais qui peut diff\u00e9rer en termes de dynamique et de r\u00e9compenses.<\/span><\/p><p><span style=\"font-weight: 400;\">Ces deux crit\u00e8res sont importants pour \u00e9valuer et comparer les performances relatives des m\u00e9thodes d\u2019apprentissage par renforcement.<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-ebacaf8 e-flex e-con-boxed e-con\" data-id=\"ebacaf8\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-e0b0f3d elementor-widget elementor-widget-heading\" data-id=\"e0b0f3d\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Quels domaines d\u2019application de l'apprentissage par renforcement ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2d1e0ac elementor-widget elementor-widget-text-editor\" data-id=\"2d1e0ac\" 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>Gr\u00e2ce \u00e0 son fort potentiel, le\u00a0Reinforcement Learning\u00a0peut bouleverser la mani\u00e8re dont les entreprises g\u00e8rent leurs activit\u00e9s op\u00e9rationnelles. \u00c0 titre d\u2019exemple, <strong>Google<\/strong>, aurait r\u00e9duit sa consommation d&rsquo;\u00e9nergie d&rsquo;environ 50 % apr\u00e8s avoir mis en \u0153uvre les technologies intelligentes d\u00e9velopp\u00e9es par\u00a0<a href=\"https:\/\/deepmind.com\/\">Deep Mind<\/a>. De m\u00eame, des start-up innovantes exploitent l&rsquo;apprentissage par renforcement\u00a0pour optimiser l\u2019utilisation des machines et des \u00e9quipements.<\/p><p>Si le Reinforcement learning est largement connu pour ses exploits en jeux vid\u00e9o, il existe une multitude de domaines d\u2019application auquel il peut apporter un v\u00e9ritable plus-value.<\/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-ad616db elementor-widget elementor-widget-heading\" data-id=\"ad616db\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-default\">La robotique<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-780940e elementor-widget elementor-widget-text-editor\" data-id=\"780940e\" 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-size: 16px; font-weight: 400;\">En robotique et en automatisation industrielle, le\u00a0<\/span><i style=\"font-size: 16px;\">Reinforcement Learning<\/i><span style=\"font-size: 16px; font-weight: 400;\">\u00a0est utilis\u00e9 pour permettre au robot de cr\u00e9er un\u00a0<\/span><span style=\"font-size: 16px; font-weight: bold;\">syst\u00e8me de contr\u00f4le adaptatif efficace\u00a0<\/span><span style=\"font-size: 16px; font-weight: 400;\">pour lui-m\u00eame qui apprend de sa propre exp\u00e9rience et de son comportement.<\/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-ec82785 elementor-widget elementor-widget-heading\" data-id=\"ec82785\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-default\">La fouille de texte<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-dfb8810 elementor-widget elementor-widget-text-editor\" data-id=\"dfb8810\" 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-size: 16px; font-weight: 400;\">Les domaines d\u2019applications du\u00a0<\/span><i style=\"font-size: 16px;\">Reinforcement Learning<\/i><span style=\"font-size: 16px; font-weight: 400;\">\u00a0comprennent \u00e9galement la fouille de texte ou\u00a0<\/span><span style=\"font-size: 16px; font-weight: bold;\">Text mining<\/span><i style=\"font-size: 16px;\">.<\/i>\u00a0<span style=\"font-size: 16px; font-weight: 400;\">\u00c0 titre d\u2019exemple, les chercheurs de Salesforce, une entreprise renomm\u00e9e de cloud computing, ont combin\u00e9 le Reinforcement Learning avec un mod\u00e8le avanc\u00e9 de g\u00e9n\u00e9ration de texte contextuel pour d\u00e9velopper un syst\u00e8me capable de produire des r\u00e9sum\u00e9s de textes longs. Pour d\u00e9couvrir une application du Text Mining version DataScientest, <a href=\"https:\/\/datascientest.com\/text-mining-appel-du-18-juin\">d\u00e9couvrez cet article !<\/a><\/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-e6feaac elementor-widget elementor-widget-heading\" data-id=\"e6feaac\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-default\">La finance<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2add591 elementor-widget elementor-widget-text-editor\" data-id=\"2add591\" 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>Si les grandes entreprises du secteur financier utilisent d\u00e9j\u00e0 <strong>les\u00a0algorithmes de Machine Learning\u00a0<\/strong>pour am\u00e9liorer leurs\u00a0<strong>activit\u00e9s de trading et d\u2019investissement en capitaux propres<\/strong>, certaines d&rsquo;entre elles ont d\u00e9j\u00e0 jet\u00e9 leur d\u00e9volu sur le\u00a0<strong>Reinforcement learning<\/strong>.\u00a0<\/p><p>Par exemple, JP Morgan a annonc\u00e9 en 2017 qu&rsquo;elle commencerait \u00e0 automatiser l&rsquo;ex\u00e9cution de gros ordres de trading. Leur mod\u00e8le, form\u00e9 sur des milliards de transactions historiques, permettrait d&rsquo;ex\u00e9cuter la transaction rapidement, \u00e0 des prix optimaux.<\/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-d700451 elementor-widget elementor-widget-heading\" data-id=\"d700451\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-default\">La sant\u00e9<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-0fec2a4 elementor-widget elementor-widget-text-editor\" data-id=\"0fec2a4\" 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-size: 16px; font-weight: 400;\">Des articles r\u00e9cents sugg\u00e8rent de multiples applications du\u00a0<\/span><span style=\"font-size: 16px; font-weight: bold;\">Reinforcement Learning<\/span>\u00a0<span style=\"font-size: 16px; font-weight: 400;\">dans le secteur des\u00a0<\/span><span style=\"font-size: 16px; font-weight: bold;\">soins de sant\u00e9<\/span><span style=\"font-size: 16px; font-weight: 400;\">. Parmi elles, on peut citer le dosage des m\u00e9dicaments, l&rsquo;optimisation des politiques de traitement pour les personnes souffrant de maladies chroniques, les essais cliniques, etc.<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-12ae5c3 e-flex e-con-boxed e-con\" data-id=\"12ae5c3\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-48c35ab elementor-widget elementor-widget-heading\" data-id=\"48c35ab\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">R\u00e9f\u00e9rences<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-a430a6b elementor-widget elementor-widget-text-editor\" data-id=\"a430a6b\" 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<ol><li><a href=\"https:\/\/arxiv.org\/pdf\/1908.06973.pdf\">Y. Li, Reinforcement Learning Applications, Ao\u00fbt 2019<\/a><\/li><li><a href=\"https:\/\/perfectial.com\/blog\/reinforcement-learning-applications\/\">M. Kovtun, Reinforcement Learning Applications: A Brief Guide on How to Get Business Value from RL<\/a><\/li><li><a href=\"https:\/\/medium.com\/@jonathan_hui\/rl-reinforcement-learning-terms-242baac11907\">J. Hui, RL \u2014 Reinforcement Learning Terms, Septembre 2017<\/a><\/li><li><a href=\"https:\/\/intellipaat.com\/blog\/supervised-learning-vs-unsupervised-learning-vs-reinforcement-learning\/\">Supervised Learning vs. Unsupervised Learning vs Reinforcement Learning, IntelliPaat, Decembre 2019<\/a><\/li><li><a href=\"https:\/\/en.wikipedia.org\/wiki\/Reinforcement_learning\">Reinforcement Learning, Wikipedia<\/a><\/li><\/ol>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-beb1f42 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"beb1f42\" 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-524c31b\" data-id=\"524c31b\" 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-db1ee8d elementor-widget elementor-widget-html\" data-id=\"db1ee8d\" data-element_type=\"widget\" data-widget_type=\"html.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<script type=\"application\/ld+json\">\r\n{\r\n  \"@context\": \"https:\/\/schema.org\",\r\n  \"@type\": \"FAQPage\",\r\n  \"mainEntity\": [{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Qu'est ce que le Reinforcement Learning ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Le Reinforcement Learning d\u00e9signe l\u2019ensemble des m\u00e9thodes qui permettent \u00e0 un agent d\u2019apprendre \u00e0 choisir quelle action prendre, et ceci de mani\u00e8re autonome.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Comment formuler un probl\u00e8me de Reinforcement learning ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"L\u2019apprentissage par renforcement n\u00e9cessite d\u2019introduire un certain nombre de concepts et de m\u00e9triques\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"En quoi le Reinforcement Learning est diff\u00e9rent des autres m\u00e9thodes d\u2019apprentissage ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"De mani\u00e8re g\u00e9n\u00e9rale, l\u2019apprentissage automatique correspond \u00e0 l\u2019apprentissage r\u00e9alis\u00e9 par des algorithmes \u00e0 partir de donn\u00e9es en vue de r\u00e9aliser des pr\u00e9dictions ou de prendre des d\u00e9cisions.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Quels les principaux challenges du Reinforcement Learning ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Le choix de l\u2019algorithme le plus appropri\u00e9 repr\u00e9sente un d\u00e9fi important \u00e9tant donn\u00e9 la grande vari\u00e9t\u00e9 d\u2019approches existantes. Celles-ci reposent sur des m\u00e9thodes diff\u00e9rentes et utilisent de nombreux hyper-param\u00e8tres. Dans ce contexte, il est essentiel de d\u00e9finir les bonnes mesures pour \u00e9valuer ses performances et de choisir avec soin ses hyper-param\u00e8tres.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Quels domaines d\u2019application de l'apprentissage par renforcement ?\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Gr\u00e2ce \u00e0 son fort potentiel, le Reinforcement Learning peut bouleverser la mani\u00e8re dont les entreprises g\u00e8rent leurs activit\u00e9s op\u00e9rationnelles. \u00c0 titre d\u2019exemple, Google, aurait r\u00e9duit sa consommation d\u2019\u00e9nergie d\u2019environ 50 % apr\u00e8s avoir mis en \u0153uvre les technologies intelligentes d\u00e9velopp\u00e9es par Deep Mind. De m\u00eame, des start-up innovantes exploitent l\u2019apprentissage par renforcement pour optimiser l\u2019utilisation des machines et des \u00e9quipements.\"\r\n    }\r\n  }]\r\n}\r\n<\/script>\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>Le Reinforcement Learning, ou apprentissage par renforcement en fran\u00e7ais, suscite depuis quelques ann\u00e9es un tr\u00e8s grand int\u00e9r\u00eat. On se souvient encore de ce moment historique o\u00f9 AlphaGo, une machine d\u2019Intelligence Artificielle, a r\u00e9ussi \u00e0 vaincre un champion humain incontest\u00e9 ! Depuis, le Reinforcement Learning a r\u00e9alis\u00e9 d\u2019importants progr\u00e8s, surpassant les m\u00e9thodes bien \u00e9tablies. N\u00e9anmoins, l\u2019utilisation [&hellip;]<\/p>\n","protected":false},"author":78,"featured_media":219662,"comment_status":"open","ping_status":"open","sticky":false,"template":"elementor_theme","format":"standard","meta":{"content-type":"","inline_featured_image":false},"categories":[2497],"tags":[],"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>Reinforcement Learning : D\u00e9finition et application<\/title>\n<meta name=\"description\" content=\"Vous entendez parler de Reinforcement Learning et vous voulez comprendre de quoi s&#039;agit-il ? Nos experts vous expliquent tout !\" \/>\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\/reinforcement-learning\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Reinforcement Learning : D\u00e9finition et application\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/reinforcement-learning\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2020-07-31T09:17:11+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-11-09T14:12:48+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Plan-de-travail-1.png\" \/>\n\t<meta property=\"og:image:width\" content=\"996\" \/>\n\t<meta property=\"og:image:height\" content=\"778\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"J\u00e9r\u00e9my Robert\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"J\u00e9r\u00e9my Robert\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/reinforcement-learning\",\"url\":\"https:\/\/datascientest.com\/reinforcement-learning\",\"name\":\"Reinforcement Learning : D\u00e9finition et application\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2020-07-31T09:17:11+00:00\",\"dateModified\":\"2023-11-09T14:12:48+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\"},\"description\":\"Vous entendez parler de Reinforcement Learning et vous voulez comprendre de quoi s'agit-il ? Nos experts vous expliquent tout !\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/reinforcement-learning\"]}]},{\"@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\/7af2361c48e24e79a8aa2360e6e74925\",\"name\":\"J\u00e9r\u00e9my Robert\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"contentUrl\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"caption\":\"J\u00e9r\u00e9my Robert\"},\"description\":\"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.\",\"sameAs\":[\"https:\/\/www.linkedin.com\/in\/jeremyrobert1\/\",\"Data Science\"],\"url\":\"https:\/\/datascientest.com\/author\/jeremyrobert\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Reinforcement Learning : D\u00e9finition et application","description":"Vous entendez parler de Reinforcement Learning et vous voulez comprendre de quoi s'agit-il ? Nos experts vous expliquent tout !","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\/reinforcement-learning","og_locale":"fr_FR","og_type":"article","og_title":"Reinforcement Learning : D\u00e9finition et application","og_url":"https:\/\/datascientest.com\/reinforcement-learning","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2020-07-31T09:17:11+00:00","article_modified_time":"2023-11-09T14:12:48+00:00","og_image":[{"width":996,"height":778,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Plan-de-travail-1.png","type":"image\/png"}],"author":"J\u00e9r\u00e9my Robert","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"J\u00e9r\u00e9my Robert","Dur\u00e9e de lecture estim\u00e9e":"8 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/reinforcement-learning","url":"https:\/\/datascientest.com\/reinforcement-learning","name":"Reinforcement Learning : D\u00e9finition et application","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2020-07-31T09:17:11+00:00","dateModified":"2023-11-09T14:12:48+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925"},"description":"Vous entendez parler de Reinforcement Learning et vous voulez comprendre de quoi s'agit-il ? Nos experts vous expliquent tout !","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/reinforcement-learning"]}]},{"@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\/7af2361c48e24e79a8aa2360e6e74925","name":"J\u00e9r\u00e9my Robert","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/datascientest.com\/#\/schema\/person\/image\/","url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","contentUrl":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","caption":"J\u00e9r\u00e9my Robert"},"description":"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.","sameAs":["https:\/\/www.linkedin.com\/in\/jeremyrobert1\/","Data Science"],"url":"https:\/\/datascientest.com\/author\/jeremyrobert"}]}},"_links":{"self":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/6629"}],"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\/78"}],"replies":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/comments?post=6629"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/6629\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/219662"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=6629"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=6629"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=6629"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}