{"376881":{"#nid":"376881","#data":{"type":"event","title":"Ph.D Defense by Matthew Plumlee","body":[{"value":"\u003Cp\u003E\u003Cstrong\u003ETitle: Fast methods for identifying high dimensional systems using \u003Cbr \/\u003Eobservations\u003C\/strong\u003E\u003C\/p\u003E\u003Cp\u003E\u003Cstrong\u003EAdvisors\u003C\/strong\u003E: Roshan Joseph Vengazhiyil and Jianjun Shi\u003Cbr \/\u003E\u003Cbr \/\u003E\u003Cstrong\u003ECommittee members:\u003C\/strong\u003E Dr. Jianjun Shi, Dr. Roshan Vengazhiyil, Dr. C.-F. \u003Cbr \/\u003EJeff Wu, Dr. Kamran Paynabar and Dr. Richard K. Archibald (Oak Ridge \u003Cbr \/\u003ENational Labs).\u003Cbr \/\u003E\u003Cbr \/\u003E\u003Cstrong\u003EDate, time, and venue\u003C\/strong\u003E: Wednesday, February 25, 2015, 11:00AM, GC 304\u003Cbr \/\u003E\u003Cbr \/\u003E\u003Cstrong\u003EThesis summary:\u003C\/strong\u003E\u003Cbr \/\u003EComputational modeling is a popular tool to understand a diverse set of \u003Cbr \/\u003Ecomplex systems. The output from a computational model depends on a set \u003Cbr \/\u003Eof parameters which are unknown to the designer, but a modeler can \u003Cbr \/\u003Eestimate them by collecting physical data. In the second chapter of this \u003Cbr \/\u003Ethesis, we study the action potential of ventricular myocytes and our \u003Cbr \/\u003Eparameter of interest is a function as opposed to a scalar or a set of \u003Cbr \/\u003Escalars. We develop a new modeling strategy to nonparametrically study \u003Cbr \/\u003Ethe functional parameter using Bayesian inference with Gaussian process \u003Cbr \/\u003Epriors. We also devise a new Markov chain Monte Carlo sampling scheme to \u003Cbr \/\u003Eaddress this unique problem.\u003Cbr \/\u003E\u003Cbr \/\u003EIn the more general case, computational simulation is expensive. \u003Cbr \/\u003EEmulators avoid the repeated use of a stochastic simulation by \u003Cbr \/\u003Eperforming a designed experiment on the computer simulation and \u003Cbr \/\u003Edeveloping a predictive distribution.\u0026nbsp; Random field models are \u003Cbr \/\u003Econsidered the standard in analysis of computer experiments, but the \u003Cbr \/\u003Ecurrent framework fails in high dimensional scenarios because of the \u003Cbr \/\u003Ecost of inference. The third chapter of this thesis shows by using a \u003Cbr \/\u003Eclass of experimental designs, the computational cost of inference from \u003Cbr \/\u003Erandom fields scales significantly better in high dimensions. Exact \u003Cbr \/\u003Eprediction and likelihood evaluation with close to half a million design \u003Cbr \/\u003Epoints is possible in seconds using only a laptop computer. Compared to \u003Cbr \/\u003Ethe more common space-filling designs, the proposed designs are shown to \u003Cbr \/\u003Ebe competitive in terms of prediction accuracy through simulation and \u003Cbr \/\u003Eanalytic results.\u003Cbr \/\u003E\u003Cbr \/\u003EThe fourth chapter of this thesis proposes a method to construct an \u003Cbr \/\u003Eemulator for a stochastic simulation. Existing emulators have focused on \u003Cbr \/\u003Eestimation of the mean of the simulation output, but this work presents \u003Cbr \/\u003Ean emulator for the distribution of the output in a nonparametric \u003Cbr \/\u003Esetting. This construction provides both an explicit distribution and a \u003Cbr \/\u003Efast sampling scheme. Beyond describing the emulator, this work \u003Cbr \/\u003Edemonstrates that the emulator\u0027s convergence rate is asymptotically rate \u003Cbr \/\u003Eoptimal among all possible emulators using the same sample size. \u0026nbsp;\u003Cbr \/\u003ELastly, the fifth chapter of this work investigates the use of a \u003Cbr \/\u003Emodified version of the above method to study patterns of defects on \u003Cbr \/\u003Eproducts. We achieve efficient inference on the defect patterns by \u003Cbr \/\u003Edeveloping a novel estimate of an inhomogeneous point process that is \u003Cbr \/\u003Eboth computationally tractable and asymptotically appealing.\u003C\/p\u003E","summary":null,"format":"limited_html"}],"field_subtitle":"","field_summary":"","field_summary_sentence":[{"value":"Fast methods for identifying high dimensional systems using observations"}],"uid":"27707","created_gmt":"2015-02-11 09:07:05","changed_gmt":"2016-10-08 01:46:15","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2015-03-26T11:30:00-04:00","event_time_end":"2015-03-26T13:30:00-04:00","event_time_end_last":"2015-03-26T13:30:00-04:00","gmt_time_start":"2015-03-26 15:30:00","gmt_time_end":"2015-03-26 17:30:00","gmt_time_end_last":"2015-03-26 17:30:00","rrule":null,"timezone":"America\/New_York"},"extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}