{"351201":{"#nid":"351201","#data":{"type":"event","title":"CSE Seminar: \u0022 Kernel Nonparametric Tests of Homogeneity, Independence and Multi-Variable Interaction\u0022 By: Arthur Greetton","body":[{"value":"\u003Cp class=\u0022p1\u0022\u003E\u003Cstrong\u003ETitle:\u003C\/strong\u003E\u003C\/p\u003E\u003Cp class=\u0022p1\u0022\u003EKernel Nonparametric Tests of Homogeneity, Independence and Multi-Variable Interaction\u003C\/p\u003E\u003Cp class=\u0022p2\u0022\u003E\u003Cstrong\u003EAbstract:\u003C\/strong\u003E\u003C\/p\u003E\u003Cp class=\u0022p1\u0022\u003EWe consider three nonparametric hypothesis testing problems: (1) Given samples\u0026nbsp;from distributions p and q, a homogeneity test determines whether to accept\u0026nbsp;or reject\u0026nbsp;p=q; (2) Given a joint distribution\u0026nbsp;pixy\u0026nbsp;over random variables x and y,\u0026nbsp;an independence test investigates whether\u0026nbsp;pixy\u0026nbsp;= p_x p_y, (3) Given a joint\u0026nbsp;distribution over several variables, we may test for whether there exist\u0026nbsp;factorization\u0026nbsp;(e.g., P_xyz = P_xyP_z, or for the case of total independence,\u0026nbsp;P_xyz=P_xP_yP_z). The final test (3) is of particular interest in fitting\u0026nbsp;directed graphical models, as it may be used in detecting cases where two\u0026nbsp;independent causes individually have weak influence on a third dependent\u0026nbsp;variable, but their combined effect has a strong influence, even when these\u0026nbsp;variables have high dimension.\u003C\/p\u003E\u003Cp class=\u0022p1\u0022\u003EWe present nonparametric tests for the three cases described, based on\u0026nbsp;distances between embeddings of probability measures to reproducing kernel\u0026nbsp;Hilbert spaces (RKHS), which constitute the test statistics (e.g.\u0026nbsp;for\u0026nbsp;independence, the distance is between the embedding of the joint, and that of\u0026nbsp;the product of the marginals). The tests benefit from decades of machine\u0026nbsp;research on kernels for various domains, and thus apply to distributions on\u0026nbsp;high dimensional vectors, images, strings, graphs, groups, and semigroups,\u0026nbsp;among others. The energy distance and distance covariance statistics are\u0026nbsp;particular instances of these RKHS statistics. Finally, the tests can be\u0026nbsp;applied for time series data, using a wild bootstrap procedure to approximate\u0026nbsp;the null hypothesis.\u003C\/p\u003E\u003Cp class=\u0022p2\u0022\u003E\u003Cstrong\u003EBio\u003C\/strong\u003E\u003C\/p\u003E\u003Cp class=\u0022p3\u0022\u003EArthur Gretton is a Reader (Associate Professor) with the Gatsby Computational\u0026nbsp;Neuroscience Unit, CSML, UCL, which he joined in 2010. He received degrees in\u0026nbsp;physics and systems engineering from the Australian National University, and a\u0026nbsp;PhD with Microsoft Research and the Signal Processing and Communications\u0026nbsp;Laboratory at the University of Cambridge. He worked from 2002-2012 at the MPI\u0026nbsp;for Biological Cybernetics, and from 2009-2010 at the Machine Learning\u0026nbsp;Department, Carnegie Mellon University.\u0026nbsp;Arthur\u0027s research interests include machine learning, kernel methods,\u0026nbsp;statistical learning theory, nonparametric hypothesis testing, blind source\u0026nbsp;separation, Gaussian processes, and non-parametric techniques for neural data\u0026nbsp;analysis. He has been an associate editor at IEEE Transactions on Pattern\u0026nbsp;Analysis and Machine Intelligence from 2009 to 2013, an Action Editor for JMLR\u0026nbsp;since April 2013, a member of the NIPS Program Committee in 2008 and 2009, an\u0026nbsp;Area Chair for ICML in 2011 and 2012, and a member of the COLT Program\u0026nbsp;Committee in 2013.\u003C\/p\u003E","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"CSE Seminar: \u0022 Kernel Nonparametric Tests of Homogeneity, Independence and Multi-Variable Interaction\u0022 By: Arthur Greetton"}],"uid":"28150","created_gmt":"2014-12-02 16:40:12","changed_gmt":"2016-10-08 01:56:28","author":"Birney Robert","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2014-12-05T13:00:00-05:00","event_time_end":"2014-12-05T14:00:00-05:00","event_time_end_last":"2014-12-05T14:00:00-05:00","gmt_time_start":"2014-12-05 18:00:00","gmt_time_end":"2014-12-05 19:00:00","gmt_time_end_last":"2014-12-05 19:00:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"47223","name":"College of Computing"},{"id":"50877","name":"School of Computational Science and Engineering"}],"categories":[],"keywords":[{"id":"111371","name":"CSE Seminar:"},{"id":"111381","name":"Independence and Multi-Variable Interaction"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1795","name":"Seminar\/Lecture\/Colloquium"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp class=\u0022p1\u0022\u003ELe Song\u003C\/p\u003E\u003Cp class=\u0022p1\u0022\u003E\u003Ca href=\u0022mailto:bdilkina@cc.gatech.edu\u0022\u003Elsong@cc.gatech.edu\u003C\/a\u003E\u003C\/p\u003E","format":"limited_html"}],"email":[],"slides":[],"orientation":[],"userdata":""}}}