Challenges in applying ranking and selection after search

摘要

It is tempting to reuse simulation replications taken during a simulation optimization search as input to a ranking-and-selection procedure, especially when generating replications is computationally expensive. Yet when a search identifies new systems based on the observed performance of explored systems, the resulting search replications are conditionally dependent given the sequence of returned systems. This dependence can mislead the selection decisions of ranking-and-selection procedures that reuse search data, in some cases to the point of violating guarantees on the probabilities of correct and good selection.