{"361231":{"#nid":"361231","#data":{"type":"event","title":"IRIM Robotics Seminar\u2013Byron Boots","body":[{"value":"\u003Cp\u003EGeorgia Tech\u2019s\u0026nbsp;Byron Boots\u0026nbsp;presents \u201cMethod of Moments for Learning Dynamical Systems\u201d as part of its Robotics Seminar Series.\u0026nbsp;The event will be held in the\u0026nbsp;TSRB Banquet Hall from 12-1 p.m. and is open to the public.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EAbstract\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EA major challenge in machine learning is to reliably and automatically discover hidden\u0026nbsp;structure in high-dimensional data. This is an especially formidable problem for\u0026nbsp;sequential data: revealing the dynamical system that governs a complex time series is\u0026nbsp;often not just difficult, but provably intractable. Popular maximum likelihood strategies\u0026nbsp;for learning dynamical system models are slow in practice and often get stuck at poor\u0026nbsp;local optima, problems that greatly limit the utility of these techniques when learning\u0026nbsp;from real-world data. Although these drawbacks were long thought to be unavoidable,\u0026nbsp;recent work has shown that progress can be made by shifting the focus of learning to\u0026nbsp;realistic instances that rule out the intractable cases.\u003C\/p\u003E\u003Cp\u003EIn this talk, I will present a new family of computational approaches for learning\u0026nbsp;dynamical system models with a particular focus on problems relevant to robotics. The key insight is that low-order moments of observed data\u0026nbsp;often possess structure that can be revealed by powerful spectral decomposition\u0026nbsp;methods, and, from this structure, model parameters can be directly recovered.\u0026nbsp;Based\u0026nbsp;on this insight, we design highly effective algorithms for learning popular parametric models like\u0026nbsp;Kalman Filters and Hidden Markov Models, as well as an expressive new class of\u0026nbsp;nonparametric models via reproducing kernels. Unlike maximum likelihood-based\u0026nbsp;approaches, these new learning algorithms are statistically consistent, computationally\u0026nbsp;efficient, and easy to implement using established matrix-algebra techniques. The\u0026nbsp;result is a powerful framework for learning dynamical system models with state-of-the-art performance on video, robotics, and biological modeling problems.\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EBio\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003EByron Boots is an assistant professor in the School of Interactive Computing at Georgia Tech. Prior to\u0026nbsp;joining Georgia Tech, he was a postdoctoral researcher working with Dieter Fox in the Robotics and State Estimation Lab at the University of Washington. He received his\u0026nbsp;Ph.D. in Machine Learning from Carnegie Mellon University in 2012, where Geoffrey Gordon was his advisor.\u0026nbsp;Boots\u2019s work on learning models of dynamical\u0026nbsp;systems received the 2010 Best Paper award at the International Conference on Machine Learning (ICML-2010). His research focuses on modeling and control\u0026nbsp;problems at the intersection of statistical machine learning, artificial intelligence, and robotics.\u003C\/p\u003E","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp class=\u0022p1\u0022\u003EGeorgia Tech\u2019s\u0026nbsp;Byron Boots\u0026nbsp;presents \u201cMethod of Moments for Learning Dynamical Systems\u201d as part of its Robotics Seminar Series.\u0026nbsp;The event will be held in the\u0026nbsp;TSRB Banquet Hall from 12-1 p.m. and is open to the public.\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Byron Boots presents \u201cMethod of Moments for Learning Dynamical Systems\u201d as part of its Robotics Seminar Series."}],"uid":"27255","created_gmt":"2015-01-06 16:32:44","changed_gmt":"2017-04-13 21:20:47","author":"Josie Giles","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2015-02-25T11:00:00-05:00","event_time_end":"2015-02-25T12:00:00-05:00","event_time_end_last":"2015-02-25T12:00:00-05:00","gmt_time_start":"2015-02-25 16:00:00","gmt_time_end":"2015-02-25 17:00:00","gmt_time_end_last":"2015-02-25 17:00:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"hg_media":{"317171":{"id":"317171","type":"image","title":"Byron Boots","body":null,"created":"1449244974","gmt_created":"2015-12-04 16:02:54","changed":"1475895024","gmt_changed":"2016-10-08 02:50:24","alt":"Byron Boots","file":{"fid":"201768","name":"byronboots.jpeg","image_path":"\/sites\/default\/files\/images\/byronboots_0.jpeg","image_full_path":"http:\/\/www.tlwarc.hg.gatech.edu\/\/sites\/default\/files\/images\/byronboots_0.jpeg","mime":"image\/jpeg","size":21673,"path_740":"http:\/\/www.tlwarc.hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/images\/byronboots_0.jpeg?itok=CQ5xiJl1"}}},"media_ids":["317171"],"related_links":[{"url":"http:\/\/www.cc.gatech.edu\/~bboots3\/","title":"Byron Boots"},{"url":"http:\/\/robotics.gatech.edu\/","title":"Center for Robotics \u0026 Intelligent Machines"}],"groups":[{"id":"47223","name":"College of Computing"},{"id":"50876","name":"School of Interactive Computing"},{"id":"142761","name":"IRIM"}],"categories":[],"keywords":[{"id":"1808","name":"graduate students"},{"id":"81491","name":"Institute for Robotics and Intelligent Machines (IRIM)"},{"id":"667","name":"robotics"},{"id":"167194","name":"seminar series"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1795","name":"Seminar\/Lecture\/Colloquium"}],"invited_audience":[{"id":"78751","name":"Undergraduate students"},{"id":"78761","name":"Faculty\/Staff"},{"id":"78771","name":"Public"},{"id":"174045","name":"Graduate students"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003EJosie Giles\u003Cbr \/\u003EIRIM Marketing Communications Mgr.\u003Cbr \/\u003E\u003Ca href=\u0022mailto:josie@gatech.edu\u0022\u003Ejosie@gatech.edu\u003C\/a\u003E\u003C\/p\u003E","format":"limited_html"}],"email":[],"slides":[],"orientation":[],"userdata":""}}}