Screening and selection procedures with control variates and correlation induction techniques

Abstract

We consider the problem of identifying the simulated system with the best expected performance measure when the number of alternatives is finite and small (often < 500). Recently, more research efforts in the simulation community have been directed to develop ranking and selection (R&S) procedures capable of exploiting variance reduction techniques (especially the control variates). In this article, we propose new R&S procedures that can jointly use control variates and correlation induction techniques (including antithetic variates and Latin hypercube sampling). Empirical results and a realistic illustration show that the proposed procedures outperform the conventional procedures using sample means or control variates alone. © 2012 Wiley Periodicals, Inc. Naval Research Logistics, 2012

Simulation optimization in security screening systems subject to budget and waiting time constraints

“…Most ranking and selection (R&S) procedures belong to this category, and they can be further classified as the frequentist‐type procedure and the Bayesian‐type procedure. The frequentist‐type often provides a statistical guarantee on selecting the best solution (e.g., Tsai et al, 2017; Tsai & Kuo, 2012). The second is the optimal computing budget allocation (OCBA) framework under the Bayesian formulation, which allocates a given number of simulation budgets to maximize the posterior probability of correct selection (e.g., Chen et al, 2010).…”

Combined variance reduction techniques in fully sequential selection procedures

“…It should also be noted that in Subsections 3.2.1–3.2.3, we assume that the conditional expectations and can be obtained analytically for every possible value of . Then, a statistically valid FSP that can efficiently employ these CV + CE combined models can be designed in a similar fashion as in Tsai and Kuo () and Tsai and Nelson (), which is presented in Appendix B.1. As discussed in Section 2.1, this assumption can be relaxed to make the combined models more widely applicable since it may not be possible to apply the CE technique in some simulation replications.…”

“…In all experiments, we set the nominal PCS = 0.95, the preliminary‐stage sample size , and the first‐stage sample size . These algorithm parameter settings are based on the guidelines provided in Tsai and Kuo () and Tsai and Nelson (). For each configuration, 500 trials of each procedure are performed to compare the performance measures, including the estimated PCS and the average number of simulated observations per system (ANS).…”

“…In addition to a sequence of papers that considered CV techniques in R&S procedures (eg, Nelson & Staum, ; Tsai & Nelson, ; Tsai et al, ), there is one more related work of Tsai and Kuo (), which further exploited models combining CV and input techniques (ie, CV + AV and CV + LHS). A summary of existing research work on VRTs is given in Table , which includes (1) comparative studies to test some of the VRTs under various experimental conditions; and (2) specifically developed R&S procedures to exploit VRTs.…”

“…A summary of existing research work on VRTs is given in Table , which includes (1) comparative studies to test some of the VRTs under various experimental conditions; and (2) specifically developed R&S procedures to exploit VRTs. Different from Tsai and Kuo (), in this paper, we consider combined models of CV with output techniques (ie, CV + CE and CV + PS), and design new fully sequential procedures when combined models may sometimes not be applicable (which motivates us to use CV + PCE). This research work is also motivated by the observation that output techniques can often obtain more statistical efficiency than input techniques, as shown in the related literature (eg, Avramidis & Wilson, ; Grant ; Sabuncuoglu et al, ).…”