Optimization of simulation experiments

• Authors: R. E. Taylor;J. W. Schmidt;V. Chachra
• Venue: WSC '73 Proceedings of the 6th conference on Winter simulation
• Year: 1973

The basic objectives of this paper are two-fold. The first objective is to illustrate the use of three multivariable optimization techniques as they are applied in an interactive fashion to the optimization of simulation experiments. The second and more important objective is to present the rationale behind the termination criterion for simulation experiments which is applicable to virtually any multivariable optimization procedure. The termination criterion is statistically based and includes cost factors prevalent for running the simulation as well as the potential savings from continued application of the search. The optimization techniques to be considered in the paper are 1. The sequential one factor-at-a-time technique as proposed by Friedman and Savage, 2. The pattern search method of Hooke and Jeeves, and 3. The successive quadratic approximation technique of Schmidt and Taylor. It is shown that the termination criteria based upon economic and statistical considerations is most effective for simulation experiments.