{"673320":{"#nid":"673320","#data":{"type":"event","title":"IC Spring Seminar Series with Guest Speaker Yilun Du","body":[{"value":"\u003Cp\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cstrong\u003E\u003Cspan\u003EAbstract\u003C\/span\u003E\u003C\/strong\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003EGenerative AI has led to stunning successes in recent years but is fundamentally limited by the amount\u0026nbsp;of\u0026nbsp;data available.\u0026nbsp;This is especially limiting in the embodied setting \u2013 where an agent must solve new tasks in new environments. In this talk, I\u2019ll introduce the idea\u0026nbsp;of\u0026nbsp;compositional generative modeling, which enables generalization beyond the training data by building complex generative models from smaller constituents. I\u2019ll first introduce the idea\u0026nbsp;of\u0026nbsp;energy-based models and illustrate how they enable compositional generative modeling. I\u2019ll then illustrate how such compositional models enable us to synthesize complex plans for unseen tasks at inference time. Finally, I\u0027ll show how such compositionality can be applied to multiple foundation models trained on various forms\u0026nbsp;of\u0026nbsp;Internet data, enabling us to construct decision-making systems that can hierarchically plan\u0026nbsp;and solve long-horizon problems in a zero-shot manner.\u003C\/span\u003E\u003C\/span\u003E\u0026nbsp;\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u0026nbsp;\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cstrong\u003E\u003Cspan\u003EBio\u003C\/span\u003E\u003C\/strong\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/p\u003E\r\n\r\n\u003Cp\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003E\u003Cspan\u003EYilun Du is a final year PhD student at MIT CSAIL advised by Leslie Kaelbling, Tomas Lozano-Perez and Joshua Tenenbaum. His research spans the fields of machine learning and robotics, with a focus on generative models. He is supported by the NSF Graduate Research Fellowship and was previously a research fellow at OpenAI, a visiting researcher at FAIR and a student researcher at Google Deepmind.\u003C\/span\u003E\u003C\/span\u003E\u0026nbsp;\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/span\u003E\u003C\/p\u003E\r\n","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETalk Title \u2014 Generalizing Beyond the Training Distribution Through Compositional Generation\u003C\/strong\u003E\u003C\/p\u003E\r\n","format":"limited_html"}],"field_summary_sentence":[{"value":"Talk Title \u2014 Generalizing Beyond the Training Distribution Through Compositional Generation"}],"uid":"36530","created_gmt":"2024-03-05 13:12:43","changed_gmt":"2024-03-05 13:12:41","author":"Nathan Deen","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2024-03-12T11:00:00-04:00","event_time_end":"2024-03-12T12:00:00-04:00","event_time_end_last":"2024-03-12T12:00:00-04:00","gmt_time_start":"2024-03-12 15:00:00","gmt_time_end":"2024-03-12 16:00:00","gmt_time_end_last":"2024-03-12 16:00:00","rrule":null,"timezone":"America\/New_York"},"location":"IC Cafe, Technology Square Research Building, 2nd Floor","extras":[],"hg_media":{"673296":{"id":"673296","type":"image","title":"Du_Yilun.png","body":null,"created":"1709644314","gmt_created":"2024-03-05 13:11:54","changed":"1709644288","gmt_changed":"2024-03-05 13:11:28","alt":"Yilun Du","file":{"fid":"256668","name":"Du_Yilun.png","image_path":"\/sites\/default\/files\/2024\/03\/05\/Du_Yilun.png","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2024\/03\/05\/Du_Yilun.png","mime":"image\/png","size":429603,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2024\/03\/05\/Du_Yilun.png?itok=zJ-03w-s"}}},"media_ids":["673296"],"groups":[{"id":"47223","name":"College of Computing"},{"id":"50876","name":"School of Interactive Computing"}],"categories":[],"keywords":[],"core_research_areas":[],"news_room_topics":[],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}