Stochastic Kriging for Simulation on Demand
<-- Return to the list
Date: 04-07-2009
Start Time:
1:00pm
End Time: 2:00pm
Speaker: Barry Nelson, NorthWestern University
Location: 303 Mudd
ABSTRACT
In contexts as diverse as automobile assembly, semiconductor manufacturing and financial risk management, we have encountered situations where there is value in being able to run extensive stochastic simulation experiments when time is plentiful so that simulation-quality results can be delivered "on demand" when time is scarce. The standard approach is to fit polynomial metamodels to the simulation output data; however, such models often provide poor global fits, do not easily support sequential experiment design, and do not provide inference about both model error and estimation error. In this talk we describe our work on stochastic kriging, an extension of kriging as used in the design and analysis of deterministic computer experiments to the stochastic simulation context.
This is joint work with Bruce Ankenman and Jeremy Staum.
BIO
Barry L. Nelson is the Charles Deering McCormick Professor of Teaching Excellence and Chair of the Department of Industrial Engineering and Management Sciences at Northwestern University. His research focus is on the design and analysis of computer simulation experiments on models of discrete-event, stochastic systems, including methodology for simulation optimization, variance reduction, metamodeling and multivariate input modeling; and applications in manufacturing, services, finance and transportation. He has published numerous papers and two books, including Discrete-Event System Simulation, 4th edition (Prentice Hall, 2005) which has been adopted by over 60 universities. Nelson is a Fellow of INFORMS. In 2006 he received the Outstanding Simulation Publication Award from the INFORMS Simulation Society for his work on simulation optimization. He has also received the Northwestern University Alumni Association Excellence in Teaching Award.
Further information is at his website.