{"50804":{"#nid":"50804","#data":{"type":"event","title":"A Sparse Signomial Model for Classification and Regression","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETITLE:\u003C\/strong\u003E A Sparse Signomial Model for Classification and\nRegression\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003ESPEAKER:\u003C\/strong\u003E Professor Myong K. (MK) Jeong\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EABSTRACT:\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003ESupport Vector Machine\n(SVM) is one of the most popular data mining tools for solving classification\nand regression problems. Due to its high prediction accuracy, SVM has been\nsuccessfully used in various fields. However, SVM has the following drawbacks.\u0026nbsp; First, it is not easy to\nget an explicit description of the discrimination (or regression) function in\nthe original input space and to make a variable selection decision in the input\nspace. Second, depending on the magnitude and numeric range of the given data\npoints, the resulting kernel matrices may be ill-conditioned, so learning\nalgorithms may be suffered from numerical instability even though data scaling\ngenerally helps to handle this kind of issues but may not be always effective.\nThird, the selection of an appropriate kernel type and its parameters can be\ncomplex while the performance of the resulting functions is heavily influenced.\u003C\/p\u003E\n\n\u003Cp\u003ETo overcome these\ndrawbacks, this talk presents the sparse\nsignomial classification and regression (SSCR) model. SSCR seek a sparse\nsignomial function by solving a linear program to minimize the weighted sum of\nthe \u003Cem\u003E\u2113\u003C\/em\u003E1-norm of the coefficient vector of the function and the \u003Cem\u003E\u2113\u003C\/em\u003E1-norm\nof violation (or loss) caused by the function. SSCR can explorevery high\ndemensional feature spaces with less sensitivity to numerical values or numeric\nranges of the given data. Moreover, this method give an explicit description of\nthe resulting function in the original input space, which can be used for\nprediction purposes as well as interpretation purposes. We present a practical\nimplementation of SSCR based on the column generation and explore some\ntheoretical properties of the proposed formulation. Computational study shows\nthat SSCR is competitive or even better performance compared to other widely\nused learning methods for classification and regression.\u003C\/p\u003E","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"A Sparse Signomial Model for Classification and Regression","format":"limited_html"}],"field_summary_sentence":[{"value":"A Sparse Signomial Model for Classification and Regression"}],"uid":"27187","created_gmt":"2010-02-03 12:30:18","changed_gmt":"2016-10-08 01:49:35","author":"Anita Race","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2010-02-12T11:00:00-05:00","event_time_end":"2010-02-12T12:00:00-05:00","event_time_end_last":"2010-02-12T12:00:00-05:00","gmt_time_start":"2010-02-12 16:00:00","gmt_time_end":"2010-02-12 17:00:00","gmt_time_end_last":"2010-02-12 17:00:00","rrule":null,"timezone":"America\/New_York"},"extras":["free_food"],"groups":[{"id":"1242","name":"School of Industrial and Systems Engineering (ISYE)"}],"categories":[],"keywords":[{"id":"167314","name":"SSCR"},{"id":"167315","name":"SVM"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1795","name":"Seminar\/Lecture\/Colloquium"}],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}