Simulation Optimization Methods Using Direct Gradients

• Activity Number: 246
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Michael Fu*+
• Companies: Smith School of Business
• Keywords: simulation ; direct gradient estimation ; perturbation analysis ; response surface methodology ; stochastic approximation ; kriging

Abstract

In stochastic simulation settings, techniques such as perturbation analysis and the likelihood ratio method allow one to obtain unbiased estimates of the gradient of output responses as a function of input parameters without having to resort to finite differences and resimulation. Recently, several new approaches have been proposed to incorporate these direction gradients into existing simulation optimization approaches such as response surface methodology, stochastic kriging, and stochastic approximation. We provide an overview of some of these approaches.