BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Talks.cam//talks.cam.ac.uk// X-WR-CALNAME:Talks.cam BEGIN:VEVENT SUMMARY:MSG Design of Experiments Seminar Series: Simulation-based Bayesia n experimental design for computationally intensive models - Xun Huan (San dia National Laboratories\; University of Michigan) DTSTART:20180620T135500Z DTEND:20180620T144500Z UID:TALK106939@talks.cam.ac.uk CONTACT:INI IT DESCRIPTION:Selecting and performing experiments that produce the most use ful data is extremely valuable in engineering and science applications wh ere experiments are costly and resources are limited. Simulation-based e xperimental design thus provides a rigorous mathematical framework to sys tematically quantify and maximize the value of experiments while leveragi ng the existing knowledge and predictive capability of an available model .  \;We are particularly interested in design settings that accommoda te nonlinear and computationally intensive models\, such as those governe d by ordinary and partial differential equations. Employing principles fr om Bayesian statistics to characterize and quantify uncertainty\, we seek experiments that maximize the expected information gain. Computing these optimal designs using conventional approaches\, however\, is generally i ntractable. Major challenges include high dimensional parameter spaces\, expensive model simulations\, and numerical approximation and optimizatio n of the expected information gain. We thus describe practical numerical methods to help overcome these obstacles\, including global sensitivity a nalysis\, surrogate modeling via polynomial chaos\, and stochastic optimi zation.  \;The overall methodology is demonstrated through the design of combustion experiments for optimal learning of chemical rate paramete rs\, and of configurations for a supersonic jet engine to obtain measurem ents most informative on turbulent flow parameters.
LOCATION:Seminar Room 1\, Newton Institute END:VEVENT END:VCALENDAR