{"342821":{"#nid":"342821","#data":{"type":"news","title":"Georgia Tech Researchers Discover Breakthrough \u0027Gaussian Cooling\u0027 Algorithm in High Dimension","body":[{"value":"\u003Cp\u003EAdvances in computing have enabled researchers to collect massive amounts of data with relative ease but have given them an even greater challenge to analyze those enormous collections of data. A single data point might have numerous features and the effort to examine patterns across a data set can reveal an exploding number of possibilities and relationships, a number that grows \u003Cem\u003Eexponentially \u003C\/em\u003Ewith each dimension.\u003C\/p\u003E\u003Cp\u003EThese problems have proved nearly intractable, requiring weeks or even months of high performance processing to handle such data sets.\u003C\/p\u003E\u003Cp\u003EBut not anymore.\u003C\/p\u003E\u003Cp\u003EResearchers at Georgia Tech\u2019s School of Computer Science have discovered an algorithm \u2013 deemed \u201cGaussian Cooling\u201d \u2013 to accurately compute the volume of convex bodies in a matter of minutes using off-the-shelf computers.\u0026nbsp;\u003C\/p\u003E\u003Cp\u003E\u201cWith this randomized algorithm, mathematicians and researchers can estimate the volume of high dimensional objects in real time,\u201d said Santosh Vempala, a professor in the School of Computer Science (SCS) who developed the new algorithm with doctoral candidate, Ben Cousins. \u201cIt can handle bodies in 100 dimensions and greater, using a new, provably correct technique.\u201d\u003C\/p\u003E\u003Cp\u003EThe problem of estimating the volume of a set is ancient. It has spawned a stream of mathematical and algorithmic ideas throughout history, starting with the Egyptians and Greeks who developed formulas in only two or three dimensions.\u003C\/p\u003E\u003Cp\u003EHowever, in spite of rapid advances in computers and algorithms and intensive study over the past 25 years, sampling and volume computation for high-dimensional sets has evaded a practical and complete solution. Research has resulted in a suite of tools to address parts of the challenge, such as analyzing and manipulating high-dimensional objects, choosing representative samples, and learning useful properties.\u003C\/p\u003E\u003Cp\u003EWith this latest advance, total volume computation is practical for the first time.\u003C\/p\u003E\u003Cp\u003E\u0022Professor Vempala and his student, Ben Cousins, made significant and surprising improvements to take the theoretical algorithm for volume computation to where we can now solve volume problems on today\u0027s computers,\u201d said Lance Fortnow, chair of SCS. \u201cTheir work brings a major tool to the algorithmic arsenal that should have applications across the sciences.\u0022\u003C\/p\u003E\u003Cp\u003EThe algorithm can be used as a tool for high-dimensional data analysis. Such data involving great numbers of parameters abounds in many fields today, including biology, neuroscience, as well as applied areas such as medical research, which may involve numerous vital statistics across many patients, or in computer security, where the algorithm can be applied to a model of information flow to estimate number of instances where data might leak.\u003C\/p\u003E\u003Cp\u003ERavi Kannan, a principal researcher at Microsoft Research India and a member of the first team to create algorithms in this field, calls the latest discovery a \u201ctour de force.\u201d\u003C\/p\u003E\u003Cp\u003E\u201cIt is the culmination of a decades-long effort by leading researchers to develop an efficient algorithm for the problem of estimating the volume of convex sets,\u201d Kannan said. \u201cI foresee many important consequences. Congratulations to Cousins and Vempala on their achievement.\u2019\u2019\u003C\/p\u003E\u003Cp\u003EGeorgia Tech\u2019s researchers have made their algorithm publicly available as a \u003Ca href=\u0022http:\/\/www.mathworks.com\/matlabcentral\/fileexchange\/43596-volume-computation-of-convex-bodies\u0022\u003EMATLAB implementation\u003C\/a\u003E and report that the method has been downloaded\u0026nbsp; hundreds of times to date. More details of the research and the researchers\u2019 paper are available \u003Ca href=\u0022http:\/\/www.cc.gatech.edu\/~bcousins\/volume.html\u0022\u003Ehere\u003C\/a\u003E.\u003C\/p\u003E","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"New algorithm provides volume estimations of convex bodies in high dimensions with general computer hardware in real time"}],"uid":"27998","created_gmt":"2014-11-07 11:36:54","changed_gmt":"2016-10-08 03:17:26","author":"Brittany Aiello","boilerplate_text":"","field_publication":"","field_article_url":"","dateline":{"date":"2014-11-08T00:00:00-05:00","iso_date":"2014-11-08T00:00:00-05:00","tz":"America\/New_York"},"extras":[],"hg_media":{"346031":{"id":"346031","type":"image","title":"Santosh and Cousins Research News","body":null,"created":"1449245670","gmt_created":"2015-12-04 16:14:30","changed":"1475895068","gmt_changed":"2016-10-08 02:51:08","alt":"Santosh and Cousins Research News","file":{"fid":"201939","name":"santosh-cousins.jpg","image_path":"\/sites\/default\/files\/images\/santosh-cousins_0.jpg","image_full_path":"http:\/\/www.tlwarc.hg.gatech.edu\/\/sites\/default\/files\/images\/santosh-cousins_0.jpg","mime":"image\/jpeg","size":117451,"path_740":"http:\/\/www.tlwarc.hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/images\/santosh-cousins_0.jpg?itok=_aS3tT39"}}},"media_ids":["346031"],"groups":[{"id":"47223","name":"College of Computing"}],"categories":[],"keywords":[{"id":"85821","name":"ben cousins"},{"id":"108931","name":"breakthrough research"},{"id":"108941","name":"gaussian cooling algorithm"},{"id":"114601","name":"Press Release"},{"id":"167405","name":"santosh vempala"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003EPhillip Taylor\u003C\/p\u003E\u003Cp\u003ENews \u0026amp; Media Relations Manager\u003C\/p\u003E\u003Cp\u003E\u003Ca href=\u0022mailto:ptaylor@cc.gatech.edu\u0022\u003Eptaylor@cc.gatech.edu\u003C\/a\u003E\u003C\/p\u003E","format":"limited_html"}],"email":["ptaylor@cc.gatech.edu"],"slides":[],"orientation":[],"userdata":""}}}