报告内容摘要

This talk studies the statistical output analysis of parallel transient simulations. It focuses on a common type of simulations, where the simulation time correlates with the output value. In this case, when the transient simulations are deployed to parallel computing architecture to increase the simulation replications in a short time, the output sequence from the parallelized computation units no longer comprise i.i.d observations. Neglecting this fact can lead to significant bias when analyzing the simulation output. In this talk, we analyze and discuss the fundamental reasons behind this phenomenon, and propose a statistical approach to provide more accurate estimations. The proposed algorithm is applicable to general terminating simulations, where the terminating event is defined by the hitting of a set of states over the course of state transitions. We have conducted thorough numerical studies to illustrate the performance of the algorithm, and applied it to analyzing the output sequence from queueing system simulation  and control chart evaluation.

报告人简介

Nan Chen is an Assistant Professor in the Department of Industrial and Systems Engineering at National University of Singapore. He obtained his B.S. degree in Automation from Tsinghua University, and M.S. degree in Computer Science, M.S. degree in Statistics, and Ph.D. degree in Industrial Engineering from University of Wisconsin-Madison. His research interests include data analytics in complex systems, simulation modeling and design, condition monitoring and degradation modeling. He is a member of INFORMS, IIE, and IEEE.

欢迎广大老师和同学们参加！