Online Appendix to: Industrial Strength COMPASS: A Comprehensive Algorithm and Software for Optimization via Simulation

The purpose of the NGA is to find promising subregions of the solution space . Each subregion has a center, which is the sample best solution in that subregion, and a handful of other solutions around the center. In the local phase, COMPASS is initialized with the center solution as the current sample best, and the surrounding solutions as the active solutions defining the most promising area. To be consistent with the GA literature, we use the term “individual” to mean a feasible solution and its “fitness” to mean the estimated expected value of the solution. The NGA described here is designed for minimization problems with integer-ordered decision variables.