Simulation-based optimization

TL;DR: The optimization methods proposed in the dissertation are adapted from the derivative-free optimization approach, which does not try to utilize or directly estimate the gradient value, and can avoid the sensitive gradient estimation process.

Abstract: Computer simulations are used extensively as models of real systems to evaluate output responses. The choice of optimal simulation parameters can lead to improved operation, but configuring them well remains a challenging problem. Simulation-based optimization is an emerging field which integrates optimization techniques into simulation analysis. The parameter calibration or optimization problem is formulated as a stochastic programming problem whose objective function is an associated measurement of an experimental simulation. Due to the complexity of the simulation, the objective function is typically (a) subject to various levels of noise, (b) not necessarily differentiable, and (c) computationally expensive to evaluate. Contemporary simulation-based optimization methods include response surface methodology, heuristic methods and stochastic approximation. Our optimization methods proposed in the dissertation are adapted from the derivative-free optimization approach, which does not try to utilize or directly estimate the gradient value. Accordingly, we can avoid the sensitive gradient estimation process. Another important feature of our methods is to use replicated samples to reduce the noise level. The idea is similar to that of the sample-path optimization method, except that we have applied Bayesian inference tools in a novel fashion to compute variable numbers of