TY - JOUR
T1 - Calibration of stochastic computer models using stochastic approximation methods
AU - Yuan, Jun
AU - Ng, Szu Hui
AU - Tsui, Kwok Leung
PY - 2013
Y1 - 2013
N2 - Computer models are widely used to simulate real processes. Within the computer model, there always exist some parameters which are unobservable in the real process but need to be specified in the model. The procedure to adjust these unknown parameters in order to fit the model to observed data and improve predictive capability is known as calibration. Practically, calibration is typically done manually. In this paper, we propose an effective and efficient algorithm based on the stochastic approximation (SA) approach that can be easily automated. We first demonstrate the feasibility of applying stochastic approximation to stochastic computer model calibration and apply it to three stochastic simulation models. We compare our proposed SA approach with another direct calibration search method, the genetic algorithm. The results indicate that our proposed SA approach performs equally as well in terms of accuracy and significantly better in terms of computational search time. We further consider the calibration parameter uncertainty in the subsequent application of the calibrated model and propose an approach to quantify it using asymptotic approximations. © 2004-2012 IEEE.
AB - Computer models are widely used to simulate real processes. Within the computer model, there always exist some parameters which are unobservable in the real process but need to be specified in the model. The procedure to adjust these unknown parameters in order to fit the model to observed data and improve predictive capability is known as calibration. Practically, calibration is typically done manually. In this paper, we propose an effective and efficient algorithm based on the stochastic approximation (SA) approach that can be easily automated. We first demonstrate the feasibility of applying stochastic approximation to stochastic computer model calibration and apply it to three stochastic simulation models. We compare our proposed SA approach with another direct calibration search method, the genetic algorithm. The results indicate that our proposed SA approach performs equally as well in terms of accuracy and significantly better in terms of computational search time. We further consider the calibration parameter uncertainty in the subsequent application of the calibrated model and propose an approach to quantify it using asymptotic approximations. © 2004-2012 IEEE.
KW - Computer model calibration
KW - parameter uncertainty
KW - stochastic approximation
KW - stochastic computer simulation
UR - http://www.scopus.com/inward/record.url?scp=84871744240&partnerID=8YFLogxK
UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-84871744240&origin=recordpage
U2 - 10.1109/TASE.2012.2199486
DO - 10.1109/TASE.2012.2199486
M3 - RGC 21 - Publication in refereed journal
SN - 1545-5955
VL - 10
SP - 171
EP - 186
JO - IEEE Transactions on Automation Science and Engineering
JF - IEEE Transactions on Automation Science and Engineering
IS - 1
M1 - 6213574
ER -