Activity Number:
|
539
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 3, 2016 : 10:30 AM to 12:20 PM
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
Abstract #318943
|
View Presentation
|
Title:
|
A Hybrid M-Estimation Method with Application in Simultaneous Tuning and Calibration for Computer Experiments
|
Author(s):
|
Gang Han* and Ao Yuan and Qizhai Li and Haiqun Lin and Thomas Santner
|
Companies:
|
Texas A&M University and Georgetown University and Chinese Academy of Sciences and Yale School of Public Health and The Ohio State University
|
Keywords:
|
M-estimation ;
computer experiments ;
hybrid Bayesian inference
|
Abstract:
|
Tuning and calibration are processes for improving the representativeness of a computer simulation code to a physical phenomenon. The corresponding statistical models (or emulators) should have both calibration and tuning parameters besides other model parameters. In practice, the calibration parameters are typically estimated by Bayesian inference, while the tuning parameters by non-Bayesian inference methods among which a reliable and robust solution is the M-estimation. In this talk, we develop a hybrid M-estimation framework and apply it to the problem of simultaneous tuning and calibration for computer experiments. We compare the proposed method with two existing approaches and illustrate our approach in a biomechanical engineering real example.
|
Authors who are presenting talks have a * after their name.