{"id":4588,"date":"2020-07-21T07:38:44","date_gmt":"2020-07-21T06:38:44","guid":{"rendered":"https:\/\/datascientest.com\/?p=4588"},"modified":"2023-11-09T14:12:55","modified_gmt":"2023-11-09T14:12:55","slug":"transfer-learning","status":"publish","type":"post","link":"https:\/\/datascientest.com\/transfer-learning","title":{"rendered":"Transfer Learning : Qu&rsquo;est-ce que c&rsquo;est ?"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"4588\" class=\"elementor elementor-4588\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-32a4715 e-flex e-con-boxed e-con\" data-id=\"32a4715\" 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-9d09d77 elementor-widget elementor-widget-heading\" data-id=\"9d09d77\" 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\">5\u00e8me et ultime volet de notre dossier consacr\u00e9 au Deep Learning, cet article vous pr\u00e9sentera en d\u00e9tail la d\u00e9finition et le fonctionnement du Transfer Learning. Si vous souhaitez revenir sur les premiers articles, en voici la liste :<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-42ba0c0 elementor-widget elementor-widget-text-editor\" data-id=\"42ba0c0\" 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<ol><li><a href=\"https:\/\/datascientest.com\/introduction-deep-learning\">Introduction \u00e0 Deep Learning<\/a><\/li><li><a href=\"https:\/\/datascientest.com\/deep-learning-reseau-de-neurones-biologiques-ou-artificiels\">R\u00e9seau de neurones: biologique ou artificiels<\/a><\/li><li><a href=\"https:\/\/datascientest.com\/convolutional-neural-network\">Convolutional Neural Network<\/a><\/li><li><a href=\"https:\/\/datascientest.com\/fonctionnement-des-reseaux-neurones\">Fonctionnement des r\u00e9seaux neurones<\/a><\/li><li>Transfer Learning<\/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-24e0e03 elementor-widget elementor-widget-text-editor\" data-id=\"24e0e03\" 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>Table des mati\u00e8res:<\/p><ol><li><a href=\"#definitiontransferlearning\">D\u00e9finitions du transfer learning<\/a><\/li><li><a href=\"#strategiestechniques\">Sur quelles strat\u00e9gies les techniques de Transfer Learning reposent ?<\/a><\/li><li><a href=\"#resolutiondeprobleme\">Comment le Transfer Learning est mis en \u0153uvre\u00a0 pour la r\u00e9solution de probl\u00e9matiques ?<\/a><\/li><li><a href=\"#pointscles\">Points cl\u00e9s \u00e0 retenir<\/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<div class=\"elementor-element elementor-element-1787561 e-flex e-con-boxed e-con\" data-id=\"1787561\" 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-cc5e7e5 elementor-widget elementor-widget-text-editor\" data-id=\"cc5e7e5\" 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;\">Andrew Ng, scientifique en chef \u00e0 <\/span><span style=\"font-weight: 400;\">Baidu<\/span><span style=\"font-weight: 400;\"> et professeur \u00e0 <\/span><span style=\"font-weight: 400;\">Stanford<\/span><span style=\"font-weight: 400;\">, a d\u00e9clar\u00e9 lors de son tr\u00e8s populaire tutoriel <\/span><span style=\"font-weight: 400;\">NIPS 2016<\/span><span style=\"font-weight: 400;\"> que l&rsquo;apprentissage par transfert sera le prochain moteur du succ\u00e8s commercial de Machine Learning. Et il n\u2019est pas le seul \u00e0 vanter le potentiel du<\/span> <b>Transfer Learning<\/b><b>\u2026<\/b><\/p><p><span style=\"font-weight: 400;\">Dans cet article, nous vous disons tout sur<strong> cette notion qui fait de plus en plus le buzz.<\/strong> DataScientest va vous aider \u00e0 r\u00e9pondre \u00e0 3 questions <strong>fondamentales<\/strong> :\u00a0<\/span><\/p><ul><li style=\"font-weight: 400;\">Qu\u2019est-ce que le Transfer Learning ?<\/li><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">Sur quelles strat\u00e9gies les techniques de <\/span><span style=\"font-weight: 400;\">Transfer Learning<\/span><span style=\"font-weight: 400;\"> reposent\u00a0?<\/span><\/li><\/ul><p><span style=\"font-weight: 400;\">Comment sont-elles mises en \u0153uvre concr\u00e8tement pour la r\u00e9solution de probl\u00e9matiques de <\/span><span style=\"font-weight: 400;\">Deep Learning<\/span><span style=\"font-weight: 400;\">\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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-c76c87f e-flex e-con-boxed e-con\" data-id=\"c76c87f\" 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-575ea19 elementor-widget elementor-widget-menu-anchor\" data-id=\"575ea19\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\nbody.elementor-page .elementor-widget-menu-anchor{margin-bottom:0}<\/style>\t\t<div id=\"definitiontransferlearning\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1129afe elementor-widget elementor-widget-heading\" data-id=\"1129afe\" 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\">Transfer Learning : d\u00e9finition<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-09f7a93 elementor-widget elementor-widget-text-editor\" data-id=\"09f7a93\" 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;\">Commen\u00e7ons d\u2019abord par d\u00e9finir ce terme de plus en plus utilis\u00e9 en <\/span><span style=\"font-weight: 400;\">Data Science<\/span><span style=\"font-weight: 400;\">.<\/span><\/p><blockquote><p><span style=\"font-weight: 400;\">Le <\/span><span style=\"font-weight: 400;\">Transfer Learning,<\/span><span style=\"font-weight: 400;\"> ou apprentissage par transfert en fran\u00e7ais, d\u00e9signe l\u2019ensemble des m\u00e9thodes qui permettent de transf\u00e9rer les connaissances acquises \u00e0 partir de la r\u00e9solution de probl\u00e8mes donn\u00e9s pour traiter un autre probl\u00e8me.\u00a0<\/span><\/p><\/blockquote><p><span style=\"font-weight: 400;\">Le <\/span><span style=\"font-weight: 400;\">Transfer Learning<\/span><span style=\"font-weight: 400;\"> a connu un grand succ\u00e8s avec l\u2019essor du <\/span><span style=\"font-weight: 400;\">Deep Learning<\/span><span style=\"font-weight: 400;\">.\u00a0 En effet, bien souvent, les mod\u00e8les utilis\u00e9s dans ce domaine n\u00e9cessitent des <\/span><b>temps de calcul \u00e9lev\u00e9s<\/b><span style=\"font-weight: 400;\"> et des <\/span><b>ressources importantes<\/b><span style=\"font-weight: 400;\">. Or, en utilisant des mod\u00e8les pr\u00e9-entra\u00een\u00e9s comme point de d\u00e9part, le <\/span><span style=\"font-weight: 400;\">Transfer Learning<\/span><span style=\"font-weight: 400;\"> permet de d\u00e9velopper rapidement des mod\u00e8les performants et r\u00e9soudre efficacement des probl\u00e8mes complexes en <a href=\"https:\/\/datascientest.com\/formation-deep-learning\/formation-computer-vision\">Computer Vision<\/a> <\/span><span style=\"font-weight: 400;\">ou <a href=\"https:\/\/datascientest.com\/formation-deep-learning\/natural-language-processing\">Natural Language Processing<\/a> <\/span><span style=\"font-weight: 400;\">(<\/span><span style=\"font-weight: 400;\">NLP)<\/span><span style=\"font-weight: 400;\">.<\/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-42dbbcd elementor-widget elementor-widget-image\" data-id=\"42dbbcd\" 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<img decoding=\"async\" width=\"1252\" height=\"386\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.28.png\" class=\"attachment-full size-full wp-image-6092\" alt=\"transfer learning\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.28.png 1252w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.28-300x92.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.28-1024x316.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.28-768x237.png 768w\" sizes=\"(max-width: 1252px) 100vw, 1252px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Figure 1 \u2013 Approche traditionnelle vs. Approche de Transfert Learning<\/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<div class=\"elementor-element elementor-element-11ed0fb elementor-widget elementor-widget-text-editor\" data-id=\"11ed0fb\" 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;\">Intuitivement, le <\/span><strong>Transfer Learning<\/strong><span style=\"font-weight: 400;\"> s\u2019inspire fortement du processus avec lequel nous apprenons.\u00a0<\/span><\/p><blockquote><p><span style=\"font-weight: 400;\">Prenons l\u2019exemple de quelqu\u2019un qui ma\u00eetrise la guitare et souhaite apprendre \u00e0 jouer au piano. Il pourra capitaliser sur ses connaissances en musique pour apprendre \u00e0 jouer un nouvel instrument. De la m\u00eame mani\u00e8re, un mod\u00e8le de reconnaissance de voiture pourra \u00eatre tr\u00e8s rapidement r\u00e9adapt\u00e9 \u00e0 la reconnaissance de camions.<\/span><\/p><\/blockquote>\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-f277945 e-flex e-con-boxed e-con\" data-id=\"f277945\" 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-2f50400 elementor-widget elementor-widget-menu-anchor\" data-id=\"2f50400\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div id=\"strategiestechniques\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-0ebdcb9 elementor-widget elementor-widget-heading\" data-id=\"0ebdcb9\" 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\">Sur quelles strat\u00e9gies les techniques de Transfer Learning reposent ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-52004f5 elementor-widget elementor-widget-text-editor\" data-id=\"52004f5\" 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><span style=\"font-weight: 400;\">Transfer Learning<\/span><span style=\"font-weight: 400;\"> repose sur une id\u00e9e simple, celle de r\u00e9-exploiter les connaissances acquises dans d\u2019autres configurations (sources) pour la r\u00e9solution d\u2019un probl\u00e8me particulier (cible). Dans ce contexte, on peut distinguer plusieurs approches selon ce que l\u2019on souhaite transf\u00e9rer, quand et comment r\u00e9aliser le transfert. Globalement, nous pouvons distinguer <\/span><b>trois types de <\/b><b>Transfer Learning<\/b><b>\u00a0<\/b><span style=\"font-weight: 400;\">:<\/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-5eb86ae elementor-widget elementor-widget-heading\" data-id=\"5eb86ae\" 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\">1. Apprentissage par transfert inductif ou Inductive Transfer Learning<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-94cd87c elementor-widget elementor-widget-text-editor\" data-id=\"94cd87c\" 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 cette configuration, les domaines source et cible sont les m\u00eames (m\u00eame donn\u00e9es), mais les t\u00e2ches source et cible sont diff\u00e9rentes mais proches. L\u2019id\u00e9e consiste alors \u00e0 <\/span><b>utiliser les mod\u00e8les existants<\/b><span style=\"font-weight: 400;\"> pour <\/span><b>r\u00e9duire de mani\u00e8re avantageuse le champ d&rsquo;application des mod\u00e8les possibles<\/b><span style=\"font-weight: 400;\"> (biais de mod\u00e8le) comme illustr\u00e9 dans le figure ci-dessous.<\/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-91c7604 elementor-widget elementor-widget-image\" data-id=\"91c7604\" 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=\"792\" height=\"370\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.33.png\" class=\"attachment-full size-full wp-image-6093\" alt=\"Inductive Transfer Learning\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.33.png 792w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.33-300x140.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.33-768x359.png 768w\" sizes=\"(max-width: 792px) 100vw, 792px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Figure 2 \u2013 Illustration de l\u2019apprentissage par transfert inductif, ou Inductive Transfer Learning<\/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<div class=\"elementor-element elementor-element-6c30491 elementor-widget elementor-widget-text-editor\" data-id=\"6c30491\" 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;\">Par exemple, il est possible d\u2019utiliser un mod\u00e8le entra\u00een\u00e9 pour la d\u00e9tection d\u2019animaux sur des images pour construire un mod\u00e8le capable d\u2019identifier des chiens.<\/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-e24494b elementor-widget elementor-widget-heading\" data-id=\"e24494b\" 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\">2. Apprentissage par transfert non supervis\u00e9 ou Unsupervised Transfer Learning<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-798783f elementor-widget elementor-widget-text-editor\" data-id=\"798783f\" 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 le cas de l\u2019apprentissage par transfert inductif, les domaines source et cible sont similaires, mais les t\u00e2ches sont diff\u00e9rentes. Toutefois, les donn\u00e9es des deux domaines ne sont pas labellis\u00e9es.<\/span><\/p><p><span style=\"font-weight: 400;\">Il est souvent plus facile d&rsquo;obtenir de grandes quantit\u00e9s de donn\u00e9es non labellis\u00e9es, \u00e0 partir de bases de donn\u00e9es et de sources sur le web par exemple, que des donn\u00e9es labellis\u00e9es. C&rsquo;est pourquoi l&rsquo;id\u00e9e d<strong>&lsquo;utiliser l&rsquo;apprentissage non supervis\u00e9 en combinaison avec le <\/strong><\/span><strong>Transfer Learning<\/strong><span style=\"font-weight: 400;\"><strong> suscite un grand int\u00e9r\u00eat<\/strong>.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">\u00c0 titre d\u2019exemple, le <\/span><span style=\"font-weight: 400;\">Self-taught clustering<\/span><span style=\"font-weight: 400;\"> est une approche qui permet de r\u00e9aliser le clustering de petites collections de donn\u00e9es cibles non labellis\u00e9es, avec l&rsquo;aide d&rsquo;une grande quantit\u00e9 de donn\u00e9es sources non labellis\u00e9es. Cette approche s\u2019av\u00e8re <strong>plus performante<\/strong> que les approches de pointe traditionnellement utilis\u00e9es, lorsque les donn\u00e9es cibles sont labellis\u00e9es de mani\u00e8re non pertinente.<\/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-e1d1671 elementor-widget elementor-widget-heading\" data-id=\"e1d1671\" 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\">3. Apprentissage par transfert transductif, ou Transductive Transfer Learning<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-3e86ff0 elementor-widget elementor-widget-text-editor\" data-id=\"3e86ff0\" 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 cette configuration, les t\u00e2ches sources et cibles sont similaires, mais les domaines correspondants sont diff\u00e9rents soit en termes de donn\u00e9es ou de distributions de probabilit\u00e9s marginales.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Par exemple, les mod\u00e8les de <\/span><span style=\"font-weight: 400;\">NLP<\/span><span style=\"font-weight: 400;\">, tels que ceux utilis\u00e9s pour l\u2019\u00e9tiquetage morpho-syntaxique de mots, Part-Of-Speech Tagger (POS Tagger) en anglais, sont g\u00e9n\u00e9ralement entra\u00een\u00e9s et test\u00e9s sur des donn\u00e9es d&rsquo;actualit\u00e9 comme celles du <\/span><span style=\"font-weight: 400;\">Wall Street Journal. <\/span><span style=\"font-weight: 400;\">Ils peuvent \u00eatre adapt\u00e9s aux donn\u00e9es issues des r\u00e9seaux sociaux dont le contenu est diff\u00e9rent mais proche de celles des journaux.<\/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-ff0f5ea e-flex e-con-boxed e-con\" data-id=\"ff0f5ea\" 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-e6b319a elementor-widget elementor-widget-menu-anchor\" data-id=\"e6b319a\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div id=\"resolutiondeprobleme\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-457abaf elementor-widget elementor-widget-heading\" data-id=\"457abaf\" 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 le Transfer Learning est mis en \u0153uvre concr\u00e8tement pour la r\u00e9solution de probl\u00e9matiques de Deep Learning ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-870bc13 elementor-widget elementor-widget-text-editor\" data-id=\"870bc13\" 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;\">Maintenant que nous avons d\u00e9fini le <\/span><strong>Transfer Learning<\/strong><span style=\"font-weight: 400;\">, int\u00e9ressons-nous \u00e0 son <strong>application aux probl\u00e9matiques de Deep Learning<\/strong>, domaine dans lequel il conna\u00eet aujourd\u2019hui un grand succ\u00e8s.<\/span><\/p><blockquote><p><span style=\"font-weight: 400;\">L\u2019utilisation des m\u00e9thodes de <\/span><span style=\"font-weight: 400;\">Transfer Learning<\/span><span style=\"font-weight: 400;\"> en <\/span><span style=\"font-weight: 400;\">Deep Learning<\/span><span style=\"font-weight: 400;\"> consiste principalement \u00e0 exploiter des r\u00e9seaux de neurones pr\u00e9-entra\u00een\u00e9s<\/span><\/p><\/blockquote><p><span style=\"font-weight: 400;\">G\u00e9n\u00e9ralement, ces mod\u00e8les correspondent \u00e0 des algorithmes tr\u00e8s performants qui ont \u00e9t\u00e9 d\u00e9velopp\u00e9s et entra\u00een\u00e9s sur de larges bases de donn\u00e9es et qui sont aujourd\u2019hui librement partag\u00e9s.<\/span><\/p><p><span style=\"font-weight: 400;\">Dans ce contexte, on peut distinguer 2 types de strat\u00e9gies\u00a0:<\/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-a03b65e elementor-widget elementor-widget-heading\" data-id=\"a03b65e\" 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\">1. Utilisation de mod\u00e8les pr\u00e9-entra\u00een\u00e9s comme extracteurs de features<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-9c2af86 elementor-widget elementor-widget-text-editor\" data-id=\"9c2af86\" 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;\">L\u2019architecture des mod\u00e8les de <\/span><span style=\"font-weight: 400;\">Deep Learning<\/span><span style=\"font-weight: 400;\"> se pr\u00e9sente tr\u00e8s souvent sous la forme d\u2019un <\/span><b>empilement de couches de neurones<\/b><span style=\"font-weight: 400;\">. Ces couches apprennent diff\u00e9rentes caract\u00e9ristiques en fonction du niveau o\u00f9 elles se situent. La derni\u00e8re couche (g\u00e9n\u00e9ralement une couche enti\u00e8rement connect\u00e9e, dans le cas de l&rsquo;apprentissage supervis\u00e9) est utilis\u00e9e pour obtenir la sortie finale. La figure ci-dessous illustre l\u2019architecture d\u2019un mod\u00e8le de <\/span><span style=\"font-weight: 400;\">Deep Learning<\/span><span style=\"font-weight: 400;\"> utilis\u00e9 pour la d\u00e9tection de chat\/chien. <\/span><b>Plus la couche se situe en profondeur, plus elle permet d\u2019extraire des <\/b><b>features<\/b><b> sp\u00e9cifiques<\/b><span style=\"font-weight: 400;\">.<\/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-45c9393 elementor-widget elementor-widget-image\" data-id=\"45c9393\" 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=\"1338\" height=\"530\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.40.png\" class=\"attachment-full size-full wp-image-6094\" alt=\"transfer learning\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.40.png 1338w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.40-300x119.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.40-1024x406.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.40-768x304.png 768w\" sizes=\"(max-width: 1338px) 100vw, 1338px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Figure 3 \u2013 Exemple de mod\u00e8le de r\u00e9seau de neurones utilis\u00e9 pour la classification d\u2019images<\/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<div class=\"elementor-element elementor-element-09c0ed2 elementor-widget elementor-widget-text-editor\" data-id=\"09c0ed2\" 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<blockquote><p><span style=\"font-weight: 400;\">L\u2019id\u00e9e est donc de r\u00e9utiliser un r\u00e9seau pr\u00e9-entra\u00een\u00e9 sans sa couche finale. Ce nouveau r\u00e9seau fonctionne alors comme un extracteur de <\/span><i><span style=\"font-weight: 400;\">features<\/span><\/i><span style=\"font-weight: 400;\"> fixes pour la r\u00e9alisation d&rsquo;autres t\u00e2ches.<\/span><\/p><\/blockquote><p><span style=\"font-weight: 400;\">Afin d\u2019illustrer cette strat\u00e9gie, prenons le cas o\u00f9 nous souhaitons <strong>cr\u00e9er un mod\u00e8le capable d\u2019identifier l\u2019esp\u00e8ce d\u2019une fleur \u00e0 partir de son image<\/strong>. Il est alors possible d\u2019utiliser les premi\u00e8res couches du mod\u00e8le de r\u00e9seau de neurones convolutif <a href=\"https:\/\/en.wikipedia.org\/wiki\/AlexNet\">AlexNet<\/a><\/span><span style=\"font-weight: 400;\">, initialement entrain\u00e9 sur la base d\u2019image <a href=\"https:\/\/en.wikipedia.org\/wiki\/ImageNet#ImageNet_Challenge\">ImageNet<\/a> <\/span><span style=\"font-weight: 400;\">pour la classification d\u2019images.<\/span><\/p><p><span style=\"font-weight: 400;\">Si l\u2019id\u00e9e para\u00eet simple, on pourrait s\u2019interroger sur les performances d\u2019une telle approche. En pratique, elle fonctionne tr\u00e8s bien comme le montre la figure ci-dessous.<\/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-fb05c3f elementor-widget elementor-widget-image\" data-id=\"fb05c3f\" 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=\"866\" height=\"382\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.46.png\" class=\"attachment-full size-full wp-image-6095\" alt=\"transfer learning\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.46.png 866w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.46-300x132.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.46-768x339.png 768w\" sizes=\"(max-width: 866px) 100vw, 866px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Figure 4 \u2013 Performance de mod\u00e8les Deep Learning sp\u00e9cifiquement entra\u00een\u00e9s sur la t\u00e2che (vert) par rapport aux mod\u00e8les pr\u00e9-entra\u00een\u00e9s (rouge, rose)<\/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<div class=\"elementor-element elementor-element-a1f45c4 elementor-widget elementor-widget-heading\" data-id=\"a1f45c4\" 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\">2. Ajustement de mod\u00e8les pr\u00e9-train\u00e9s<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-f215173 elementor-widget elementor-widget-text-editor\" data-id=\"f215173\" 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;\">Il s&rsquo;agit d&rsquo;une <strong>technique plus complexe<\/strong>, dans laquelle non seulement la derni\u00e8re couche est remplac\u00e9e pour r\u00e9aliser la classification ou la r\u00e9gression, mais d\u2019autres couches sont \u00e9galement r\u00e9-entra\u00eener de mani\u00e8re s\u00e9lective. En effet, l<strong>es r\u00e9seaux neuronaux profonds sont des architectures hautement configurables<\/strong> avec divers hyperparam\u00e8tres. De plus, alors que les premi\u00e8res couches capturent les caract\u00e9ristiques g\u00e9n\u00e9riques, les derni\u00e8res couches se concentrent davantage sur la t\u00e2che sp\u00e9cifique \u00e0 accomplir, comme le montre la figure-ci-dessous.<\/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-10ff039 elementor-widget elementor-widget-image\" data-id=\"10ff039\" 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=\"1350\" height=\"712\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.52.png\" class=\"attachment-full size-full wp-image-6096\" alt=\"mod\u00e8le de r\u00e9seau de neurone\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.52.png 1350w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.52-300x158.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.52-1024x540.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.52-768x405.png 768w\" sizes=\"(max-width: 1350px) 100vw, 1350px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Figure 5 \u2013 Exemple de mod\u00e8le de r\u00e9seau de neurones utilis\u00e9 pour reconnaissance faciale<\/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<div class=\"elementor-element elementor-element-def8843 elementor-widget elementor-widget-text-editor\" data-id=\"def8843\" 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<blockquote><p><span style=\"font-weight: 400;\">L\u2019id\u00e9e est donc de geler (c\u2019est-\u00e0-dire fixer les poids) de certaines couches pendant l\u2019entra\u00eenement et affiner le reste pour r\u00e9pondre \u00e0 la probl\u00e9matique.\u00a0<\/span><\/p><\/blockquote><p><span style=\"font-weight: 400;\">Cette strat\u00e9gie permet de r\u00e9utiliser les connaissances en termes d&rsquo;architecture globale du r\u00e9seau et d\u2019exploiter ses \u00e9tats comme point de d\u00e9part pour l\u2019entra\u00eenement. E<\/span><b>lle permet donc d&rsquo;obtenir de meilleures performances avec un temps d\u2019entra\u00eenement plus court.<\/b><\/p><p><span style=\"font-weight: 400;\">La figure ci-dessous r\u00e9capitule les principales approches de <\/span><span style=\"font-weight: 400;\">Transfer Learning<\/span><span style=\"font-weight: 400;\"> utilis\u00e9es commun\u00e9ment en <\/span><span style=\"font-weight: 400;\">Deep Learning<\/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-78e1dfe elementor-widget elementor-widget-image\" data-id=\"78e1dfe\" 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=\"1340\" height=\"840\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.58.png\" class=\"attachment-full size-full wp-image-6097\" alt=\"transfer learning\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.58.png 1340w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.58-300x188.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.58-1024x642.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/Capture-d\u2019e\u0301cran-2020-07-21-a\u0300-08.45.58-768x481.png 768w\" sizes=\"(max-width: 1340px) 100vw, 1340px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Figure 6 \u2013 Approche de Transfer Learning en Deep Learning, <\/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<div class=\"elementor-element elementor-element-66ed8ec elementor-widget elementor-widget-text-editor\" data-id=\"66ed8ec\" 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;une des <strong>exigences fondamentales<\/strong> de l&rsquo;apprentissage par transfert <strong>est la pr\u00e9sence de mod\u00e8les<\/strong> qui fonctionnent bien sur les t\u00e2ches sources. Heureusement, de nombreuses architectures de Deep Learning de pointe sont aujourd\u2019hui partag\u00e9es librement par leurs \u00e9quipes respectives. Celles-ci s&rsquo;\u00e9tendent sur diff\u00e9rents domaines tels que la vision par ordinateur ou Computer Vision, et le Traitement du langage nature ou le NLP, deux domaines tr\u00e8s populaires en Deep Learning.<\/p><p>Parmi les mod\u00e8les tr\u00e8s fr\u00e9quemment utilis\u00e9s :<\/p><ul><li>Computer Vision : <a href=\"https:\/\/www.kaggle.com\/keras\/vgg16\/home\">VGG-16<\/a>, <a href=\"https:\/\/www.kaggle.com\/keras\/vgg19\/home\">VGG-19<\/a>, <a href=\"https:\/\/www.kaggle.com\/keras\/resnet50\">ResNet-50<\/a><\/li><li>NLP : <a href=\"https:\/\/en.wikipedia.org\/wiki\/Word2vec\">Word2Vec<\/a>, <a href=\"https:\/\/nlp.stanford.edu\/projects\/glove\/\">GloVe<\/a><\/li><\/ul>\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-069ff20 e-flex e-con-boxed e-con\" data-id=\"069ff20\" 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-0ec9d10 elementor-widget elementor-widget-menu-anchor\" data-id=\"0ec9d10\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div id=\"pointscles\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-5579b78 elementor-widget elementor-widget-heading\" data-id=\"5579b78\" 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\">Points cl\u00e9s \u00e0 retenir<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-8e2364e elementor-widget elementor-widget-text-editor\" data-id=\"8e2364e\" 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><span style=\"font-weight: 400;\">Transfer Learning<\/span><span style=\"font-weight: 400;\"> correspond \u00e0 la capacit\u00e9 \u00e0 utiliser des connaissances existantes, d\u00e9velopp\u00e9es pour la r\u00e9solution de probl\u00e9matiques donn\u00e9es, pour r\u00e9soudre une nouvelle probl\u00e9matique.<\/span><\/p><p><span style=\"font-weight: 400;\">Nous avons abord\u00e9 plusieurs approches de <\/span><span style=\"font-weight: 400;\">Transfert Learning. <\/span><span style=\"font-weight: 400;\">Afin de pouvoir choisir la strat\u00e9gie la plus adapt\u00e9e, il est important d\u2019examiner les trois questions suivantes\u00a0:\u00a0<\/span><\/p><ol><li style=\"list-style-type: none;\"><ol><li style=\"font-weight: 400;\"><b>Que veut-on transf\u00e9rer\u00a0?<\/b><span style=\"font-weight: 400;\"> C&rsquo;est la premi\u00e8re question \u00e0 se poser et la plus importante. Elle vise \u00e0 d\u00e9finir quelle partie de la connaissance est \u00e0 transf\u00e9rer de la source \u00e0 la cible, afin de r\u00e9pondre \u00e0 la probl\u00e9matique de la t\u00e2che cible.\u00a0<\/span><\/li><li style=\"font-weight: 400;\"><b>Quand veut-on transf\u00e9rer\u00a0?<\/b><span style=\"font-weight: 400;\"> Il peut y avoir des sc\u00e9narios o\u00f9 l\u2019application de m\u00e9thodes de <\/span><span style=\"font-weight: 400;\">Transfert Learning<\/span><span style=\"font-weight: 400;\"> peut conduire \u00e0 une d\u00e9gradation des r\u00e9sultats. En r\u00e9alit\u00e9, ces performances d\u00e9pendent de la similarit\u00e9 entre les domaines et les t\u00e2ches cible et source.<\/span><\/li><li style=\"font-weight: 400;\"><b>Comment va-t-on r\u00e9aliser le transfert\u00a0?<\/b><span style=\"font-weight: 400;\"> Une fois que nous avons r\u00e9pondu \u00e0 la question \u00ab\u00a0quoi\u00a0\u00bb et \u00ab\u00a0quand\u00a0\u00bb, nous pouvons commencer \u00e0 identifier la technique de <\/span><span style=\"font-weight: 400;\">Transfer Leaning<\/span><span style=\"font-weight: 400;\"> la plus adapt\u00e9e au probl\u00e8me que nous souhaitons r\u00e9soudre.\u00a0<\/span><\/li><\/ol><\/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<div class=\"elementor-element elementor-element-d12693e e-flex e-con-boxed e-con\" data-id=\"d12693e\" 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-d0dc103 elementor-widget elementor-widget-text-editor\" data-id=\"d0dc103\" 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;\"><strong>Bravo ! Vous avez termin\u00e9 notre dossier Deep Learning ! Vous savez maintenant l&rsquo;essentiel sur ses applications \u00e0 succ\u00e8s, le fonctionnement des r\u00e9seaux de neurones, les CNN et d\u00e9sormais le Transfer Learning . Vous souhaitez aller plus loin dans votre apprentissage du Deep Learning ? D\u00e9couvrez (et d\u00e9marrez prochainement) l&rsquo;une de nos formations sp\u00e9cialis\u00e9es.<\/strong><\/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-0e4086b elementor-widget__width-initial elementor-widget elementor-widget-video\" data-id=\"0e4086b\" data-element_type=\"widget\" data-settings=\"{&quot;youtube_url&quot;:&quot;https:\\\/\\\/youtu.be\\\/Y3Z5W9W_L9E&quot;,&quot;video_type&quot;:&quot;youtube&quot;,&quot;controls&quot;:&quot;yes&quot;}\" data-widget_type=\"video.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-video .elementor-widget-container{overflow:hidden;transform:translateZ(0)}.elementor-widget-video .elementor-wrapper{aspect-ratio:var(--video-aspect-ratio)}.elementor-widget-video .elementor-wrapper iframe,.elementor-widget-video .elementor-wrapper video{height:100%;width:100%;display:flex;border:none;background-color:#000}@supports not (aspect-ratio:1\/1){.elementor-widget-video .elementor-wrapper{position:relative;overflow:hidden;height:0;padding-bottom:calc(100% \/ var(--video-aspect-ratio))}.elementor-widget-video .elementor-wrapper iframe,.elementor-widget-video .elementor-wrapper video{position:absolute;top:0;right:0;bottom:0;left:0}}.elementor-widget-video .elementor-open-inline .elementor-custom-embed-image-overlay{position:absolute;top:0;right:0;bottom:0;left:0;background-size:cover;background-position:50%}.elementor-widget-video .elementor-custom-embed-image-overlay{cursor:pointer;text-align:center}.elementor-widget-video .elementor-custom-embed-image-overlay:hover .elementor-custom-embed-play i{opacity:1}.elementor-widget-video .elementor-custom-embed-image-overlay img{display:block;width:100%;aspect-ratio:var(--video-aspect-ratio);-o-object-fit:cover;object-fit:cover;-o-object-position:center center;object-position:center center}@supports not (aspect-ratio:1\/1){.elementor-widget-video .elementor-custom-embed-image-overlay{position:relative;overflow:hidden;height:0;padding-bottom:calc(100% \/ var(--video-aspect-ratio))}.elementor-widget-video .elementor-custom-embed-image-overlay img{position:absolute;top:0;right:0;bottom:0;left:0}}.elementor-widget-video .e-hosted-video .elementor-video{-o-object-fit:cover;object-fit:cover}.e-con-inner>.elementor-widget-video,.e-con>.elementor-widget-video{width:var(--container-widget-width);--flex-grow:var(--container-widget-flex-grow)}<\/style>\t\t<div class=\"elementor-wrapper elementor-open-inline\">\n\t\t\t<div class=\"elementor-video\"><\/div>\t\t<\/div>\n\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-036a674 e-flex e-con-boxed e-con\" data-id=\"036a674\" 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-dc465a5 elementor-widget elementor-widget-text-editor\" data-id=\"dc465a5\" 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><strong>Article 1\u00a0:<\/strong> <a href=\"https:\/\/datascientest.com\/introduction-deep-learning\">Introduction au deep learning et aux r\u00e9seaux de neurones<\/a><\/p><p><strong>Article 2 :<\/strong>\u00a0<a href=\"https:\/\/datascientest.com\/convolutional-neural-network\">Convolutional neural network<\/a><\/p><p><strong>Article 3 :<\/strong> <a href=\"https:\/\/datascientest.com\/fonctionnement-des-reseaux-neurones\">Fonctionnement des r\u00e9seaux de neurones<\/a><\/p><p><strong>Article 4 :<\/strong> Transfer learning\u00a0<\/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-01b28ad elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"01b28ad\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-deep-learning\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">D\u00e9couvrir notre formation en Deep Learning<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\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\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>5\u00e8me et ultime volet de notre dossier consacr\u00e9 au Deep Learning, cet article vous pr\u00e9sentera en d\u00e9tail la d\u00e9finition et le fonctionnement du Transfer Learning. Si vous souhaitez revenir sur les premiers articles, en voici la liste : Introduction \u00e0 Deep Learning R\u00e9seau de neurones: biologique ou artificiels Convolutional Neural Network Fonctionnement des r\u00e9seaux neurones [&hellip;]<\/p>\n","protected":false},"author":78,"featured_media":23761,"comment_status":"open","ping_status":"closed","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>Transfer Learning : Qu&#039;est-ce que c&#039;est ?<\/title>\n<meta name=\"description\" content=\"Envie de comprendre et ma\u00eetriser tous les concepts cl\u00e9s du Deep Learning ? DataScientest vous explique le Transfer Learning !\" \/>\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\/transfer-learning\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Transfer Learning : Qu&#039;est-ce que c&#039;est ?\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/transfer-learning\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2020-07-21T06:38:44+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-11-09T14:12:55+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/illu_DLtransfer_blog-33.png\" \/>\n\t<meta property=\"og:image:width\" content=\"2560\" \/>\n\t<meta property=\"og:image:height\" content=\"1405\" \/>\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=\"9 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/transfer-learning\",\"url\":\"https:\/\/datascientest.com\/transfer-learning\",\"name\":\"Transfer Learning : Qu'est-ce que c'est ?\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2020-07-21T06:38:44+00:00\",\"dateModified\":\"2023-11-09T14:12:55+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\"},\"description\":\"Envie de comprendre et ma\u00eetriser tous les concepts cl\u00e9s du Deep Learning ? DataScientest vous explique le Transfer Learning !\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/transfer-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":"Transfer Learning : Qu'est-ce que c'est ?","description":"Envie de comprendre et ma\u00eetriser tous les concepts cl\u00e9s du Deep Learning ? DataScientest vous explique le Transfer Learning !","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\/transfer-learning","og_locale":"fr_FR","og_type":"article","og_title":"Transfer Learning : Qu'est-ce que c'est ?","og_url":"https:\/\/datascientest.com\/transfer-learning","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2020-07-21T06:38:44+00:00","article_modified_time":"2023-11-09T14:12:55+00:00","og_image":[{"width":2560,"height":1405,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/07\/illu_DLtransfer_blog-33.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":"9 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/transfer-learning","url":"https:\/\/datascientest.com\/transfer-learning","name":"Transfer Learning : Qu'est-ce que c'est ?","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2020-07-21T06:38:44+00:00","dateModified":"2023-11-09T14:12:55+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925"},"description":"Envie de comprendre et ma\u00eetriser tous les concepts cl\u00e9s du Deep Learning ? DataScientest vous explique le Transfer Learning !","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/transfer-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\/4588"}],"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=4588"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/4588\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/23761"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=4588"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=4588"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=4588"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}