{"id":130368,"date":"2021-11-01T15:44:49","date_gmt":"2021-11-01T14:44:49","guid":{"rendered":"https:\/\/www.pauljorion.com\/blog\/?p=130368"},"modified":"2021-11-01T15:45:37","modified_gmt":"2021-11-01T14:45:37","slug":"lintelligence-incarnee-par-lapprentissage-et-levolution-par-agrim-gupta-silvio-savarese-surya-ganguli-li-fei-fei","status":"publish","type":"post","link":"https:\/\/www.pauljorion.com\/blog\/2021\/11\/01\/lintelligence-incarnee-par-lapprentissage-et-levolution-par-agrim-gupta-silvio-savarese-surya-ganguli-li-fei-fei\/","title":{"rendered":"<b>L&rsquo;intelligence incarn\u00e9e par l&rsquo;apprentissage et l&rsquo;\u00e9volution<\/b>, par Agrim Gupta, Silvio Savarese, Surya Ganguli &#038; Li Fei-Fei"},"content":{"rendered":"<p><iframe loading=\"lazy\" width=\"700\" height=\"450\" class=\"aligncenter\" src=\"https:\/\/www.youtube.com\/embed\/JPFhqrN9DlE\" title=\"YouTube video player\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture\" allowfullscreen><\/iframe><\/p>\n<p><a href=\"https:\/\/www.nature.com\/articles\/s41467-021-25874-z.pdf\" rel=\"noopener\" target=\"_blank\">Embodied intelligence via learning and evolution<\/a><\/p>\n<p>Agrim Gupta, Silvio Savarese, Surya Ganguli &#038; Li Fei-Fei<br \/>\n<em>Nature Communications<\/em> volume 12, Article number: 5721 (2021)<br \/>\n<!--more--><\/p>\n<p>L&rsquo;imbrication des processus d&rsquo;apprentissage et d&rsquo;\u00e9volution dans des niches environnementales complexes a donn\u00e9 lieu \u00e0 une remarquable diversit\u00e9 de formes morphologiques. En outre, de nombreux aspects de l&rsquo;intelligence animale sont profond\u00e9ment incarn\u00e9s dans ces morphologies \u00e9volu\u00e9es. Cependant, les principes r\u00e9gissant les relations entre la complexit\u00e9 de l&rsquo;environnement, la morphologie \u00e9volu\u00e9e et la capacit\u00e9 d&rsquo;apprentissage du contr\u00f4le intelligent restent insaisissables, car la r\u00e9alisation d&rsquo;exp\u00e9riences in silico \u00e0 grande \u00e9chelle sur l&rsquo;\u00e9volution et l&rsquo;apprentissage est difficile. Nous pr\u00e9sentons ici Deep Evolutionary Reinforcement Learning (DERL) : un cadre de calcul capable de faire \u00e9voluer diverses morphologies d&rsquo;agents pour apprendre des t\u00e2ches difficiles de locomotion et de manipulation dans des environnements complexes. En utilisant DERL, nous d\u00e9montrons plusieurs relations entre la complexit\u00e9 de l&rsquo;environnement, l&rsquo;intelligence morphologique et la capacit\u00e9 d&rsquo;apprentissage du contr\u00f4le. Premi\u00e8rement, la complexit\u00e9 de l&rsquo;environnement favorise l&rsquo;\u00e9volution de l&rsquo;intelligence morphologique, quantifi\u00e9e par la capacit\u00e9 d&rsquo;une morphologie \u00e0 faciliter l&rsquo;apprentissage de nouvelles t\u00e2ches. Deuxi\u00e8mement, nous d\u00e9montrons un effet Baldwin morphologique, c&rsquo;est-\u00e0-dire que dans nos simulations, l&rsquo;\u00e9volution s\u00e9lectionne rapidement les morphologies qui apprennent plus vite, permettant ainsi aux comportements appris tard dans la vie des premiers anc\u00eatres de s&rsquo;exprimer t\u00f4t dans la vie des descendants. Troisi\u00e8mement, nous sugg\u00e9rons une base m\u00e9caniste pour les relations ci-dessus par l&rsquo;\u00e9volution de morphologies qui sont plus stables physiquement et plus efficaces \u00e9nerg\u00e9tiquement, et qui peuvent donc faciliter l&rsquo;apprentissage et le contr\u00f4le.<\/p>\n","protected":false},"excerpt":{"rendered":"<p><iframe loading=\"lazy\" width=\"700\" height=\"450\" class=\"aligncenter\" src=\"https:\/\/www.youtube.com\/embed\/JPFhqrN9DlE\" title=\"YouTube video player\" frameborder=\"0\" allow=\"accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture\" allowfullscreen><\/iframe><\/p>\n<p><a href=\"https:\/\/www.nature.com\/articles\/s41467-021-25874-z.pdf\" rel=\"noopener\" target=\"_blank\">Embodied intelligence via learning and evolution<\/a><\/p>\n<p>Agrim Gupta, Silvio Savarese, Surya Ganguli &#038; Li Fei-Fei<br \/> <em>Nature Communications<\/em> volume 12, Article number: 5721 (2021) <\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_crdt_document":"","footnotes":""},"categories":[5833,13],"tags":[3277,8423],"class_list":["post-130368","post","type-post","status-publish","format-standard","hentry","category-ethologie","category-intelligence-artificielle","tag-evolution","tag-intelligence"],"_links":{"self":[{"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/posts\/130368","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/comments?post=130368"}],"version-history":[{"count":2,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/posts\/130368\/revisions"}],"predecessor-version":[{"id":130370,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/posts\/130368\/revisions\/130370"}],"wp:attachment":[{"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/media?parent=130368"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/categories?post=130368"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.pauljorion.com\/blog\/wp-json\/wp\/v2\/tags?post=130368"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}