Efficient Input Uncertainty Quantification Via Green Simulation Using Sample Path Likelihood Ratios

Bootstrapping is a popular tool for quantifying input uncertainty, inflated uncertainty in the simulation output caused by finite-sample estimation error in the input models. Typical bootstrap-based procedures have a nested simulation structure that requires B × R simulation runs: the outer loop bootstraps B input distributions, each of which requires R inner simulation runs. In this article, we present a measure-theoretic framework for constructing a sample path likelihood ratio and propose an efficient input uncertainty quantification procedure using two green simulation estimators. The proposed procedures reuse the same R inner simulation outputs in all outer loops by reweighting them using appropriately defined likelihood ratios. Both procedures produce asymptotically valid confidence intervals for the expected simulation output under the true input distribution. Our numerical results show that the proposed procedures have efficiency gains compared to other popular bootstrap-based alternatives.

[1]  Jeremy Staum,et al.  Green Simulation , 2017, ACM Trans. Model. Comput. Simul..

[2]  Jeremy Staum,et al.  Green simulation designs for repeated experiments , 2015, 2015 Winter Simulation Conference (WSC).

[3]  Lee W. Schruben,et al.  Uniform and bootstrap resampling of empirical distributions , 1993, WSC '93.

[4]  Wei Xie,et al.  A Bayesian Framework for Quantifying Uncertainty in Stochastic Simulation , 2014, Oper. Res..

[5]  P. L’Ecuyer.  Two approaches for estimating the gradient in functional form , 1993, WSC '93.

[6]  P. Frazier,et al.  Advanced tutorial: Input uncertainty and robust analysis in stochastic simulation , 2016, 2016 Winter Simulation Conference (WSC).

[7]  Lee W. Schruben,et al.  Resampling methods for input modeling , 2001, Proceeding of the 2001 Winter Simulation Conference (Cat. No.01CH37304).

[8]  Barry L. Nelson,et al.  A quicker assessment of input uncertainty , 2013, 2013 Winter Simulations Conference (WSC).

[9]  R. R. Barton,et al.  Tutorial: Input uncertainty in outout analysis , 2012, Proceedings Title: Proceedings of the 2012 Winter Simulation Conference (WSC).

[10]  P. Glasserman,et al.  Robust risk measurement and model risk , 2012 .

[11]  David J. Eckman,et al.  GREEN SIMULATION OPTIMIZATION USING LIKELIHOOD RATIO ESTIMATORS , 2018, 2018 Winter Simulation Conference (WSC).

[12]  Barry L. Nelson,et al.  Quickly Assessing Contributions to Input Uncertainty , 2015 .

[14]  Enlu Zhou,et al.  ONLINE QUANTIFICATION OF INPUT UNCERTAINTY FOR PARAMETRIC MODELS , 2018, 2018 Winter Simulation Conference (WSC).

[15]  Leonidas J. Guibas,et al.  Optimally combining sampling techniques for Monte Carlo rendering , 1995, SIGGRAPH.

[16]  Barry L. Nelson,et al.  UNBIASED METAMODELING VIA LIKELIHOOD RATIOS , 2018, 2018 Winter Simulation Conference (WSC).

[17]  Jason H. Goodfriend,et al.  Discrete Event Systems: Sensitivity Analysis and Stochastic Optimization by the Score Function Method , 1995 .

[18]  Henry Lam,et al.  SUBSAMPLING VARIANCE FOR INPUT UNCERTAINTY QUANTIFICATION , 2018, 2018 Winter Simulation Conference (WSC).

[19]  Jean-Marie Cornuet,et al.  Adaptive Multiple Importance Sampling , 2009, 0907.1254.

[20]  R. Cheng,et al.  Sensitivity of computer simulation experiments to errors in input data , 1997 .

[21]  Barry L. Nelson,et al.  Advanced tutorial: Input uncertainty quantification , 2014, Proceedings of the Winter Simulation Conference 2014.

[22]  M. D. McKay,et al.  Monte Carlo estimation under different distributions using the same simulation , 1987 .

[23]  Barry L. Nelson,et al.  Single-experiment input uncertainty , 2015, J. Simulation.

[24]  A. Owen,et al.  Safe and Effective Importance Sampling , 2000 .