Abstract
We consider a Bayesian ranking and selection problem in the presence of input distribution uncertainty. The distribution uncertainty is treated from a robust perspective. A naive extension of the knowledge gradient (KG) policy fails to converge in the new robust setting. We propose several stationary policies that extend KG in various aspects. Numerical experiments show that the proposed policies have excellent performance in terms of both probability of correction selection and normalized opportunity cost.
Xiaowei Zhang
Associate Professor
My research research focuses on methodological advances in stochastic simulation and optimization, decision analytics, and reinforcement learning, with applications in service operations management, financial technology, and digital economy.