You are here: Home NEWS & EVENTS Events Calendar
Events Calendar
  • [April 5, 2017]

    Simulation Optimization: Efficient Sampling Procedure for Finding the Best

  • Speaker:
    Yijie Peng
    Wednesday, April 5, 2017
    Room 512, Founder Building
    Leyuan Shi
  • Abstract
  • Simulation optimization is a research area that aims to efficiently combine simulation with optimization. This talk focuses two actively studied areas in simulation optimization: ranking and selection (R&S) and sensitivity analysis. The speaker shows an unexpected phenomenon that many popular sampling allocation procedures in R&S decrease the probability of correct selection (PCS), which is caused by the imperfect theoretical foundation of these methods. To address the problem, a dynamic sampling and selection framework is proposed. Under this framework, R&S can be formulated as a stochastic control problem and efficiently solved in an approximate dynamic programming (ADP) paradigm. Two ADP approaches are provided. One using a single feature of the value function sequentially achieves an asymptotically optimal sampling ratio that cannot be achieved by the many existing sequential sampling procedures. Another ADP approach using two features avoids the non-monotonicity of the PCS. In sensitivity analysis, the speaker introduces a new unbiased stochastic derivative estimator called generalized likelihood ratio (GLR) method that can handle a large scope of discontinuities. Several applications are put together under the umbrella of distribution sensitivities and solved uniformly by GLR.
  • Biography
  • Yijie Peng received his Ph.D. from the Management Science department of Fudan University. Dr. Peng is currently a Research Assistant Professor of George Mason University (GMU). Before joining GMU, he worked as a postdoctoral scholar at Fudan University and R.H. Smith School of Business at University of Maryland at College Park. Many of his publications appear in high-quality journals including IEEE Transaction on Automatic Control, INFORMS Journal on Computing, and Quantitative Finance. He was awarded with the honor of the outstanding Ph.D. thesis by Fudan University in 2014, and won the best paper award from Shanghai Forum on Operations Research and Control Theory in 2016.