Convergence of a Single Run Simulation Optimization Algorithm

Conventional methods for simulation optimization may require a large number of simulation runs and are thus computationally expensive. To reduce computational effort, we have previously conducted experimental investigations of single run optimization algorithms which estimate the optimum in a single simulation run. Our earlier studies for queueing systems have focused on empirical results. In this study, we apply a single run optimization algorithm to a simple optimization problem related to an autoregressive process of order one. It is shown that the algorithm converges to the true optimum with probability one. This provides a first step towards analytical understanding of such single run simulation optimization algorithms.