ORIE’s Jing Xie wins INFORMS Computing Society’s best student paper award
Using a computer to simulate a complex system, such as ambulance services in a large city or a manufacturing production line, is an excellent way to understand the implications of particular operational choices, such as where to locate the ambulances or whether to operate the line on a given day. When the objective is to determine whether a decision meets a predetermined criterion, such as answering 70% of ambulance calls on time or achieving a certain revenue goal, it takes many simulation runs to check all of the operational choices, with each choice requiring multiple runs because of the inherent randomness of demand. Especially if computing resources are purchased and paid for on an as-needed basis (such as from the Amazon Elastic Compute Cloud, finding a better way to allocate computing resources can result in significant cost savings.
Reducing simulation effort
Fifth-year ORIE Ph.D. student Jing Xie has found a tractable way to reduce the required number of simulation runs. At the recent meeting of the Institute for Operations Research and the Management Sciences (INFORMS), the INFORMS Computing Society (ICS) awarded its 2013 Student Paper Award to Sequential Bayes-Optimal Policies for Multiple Comparisons with a Known Standard, a paper describing this approach that she coauthored with her advisor, Assistant Professor Peter Frazier.
In presenting the award, selection committee member Michael Fu, Ralph J. Tyser Professor of Management Science at the University of Maryland, said “overall, the paper is well written and makes important contributions to both the theory and practice of simulation optimization by using a rigorous modeling framework that leads to useful implementable algorithms.
“I was excited to receive this award,” Xie said. “It is a great honor.” She is grateful to Professor Frazier for his efforts as coauthor, particularly as the paper went through successive revisions. The paper has been accepted for publication in the journal Operations Research.
The paper develops a method for determining how best to allocate simulation resources among the alternatives being considered, using information from the simulation runs completed so far to guide the choice of the next alternatives to simulate. The authors prove that their method uses this information optimally among all such learning approaches that are built upon a formulation first explored by the Reverend Thomas Bayes (1702 -1761).
Further, using the ambulance and production line examples described earlier, the authors show that their allocation method is substantially better than methods described in the literature. In particular choosing, at random, the next alternative to simulate (which corresponds to giving equal weight to all alternatives) can require an order of magnitude more computing resources than their approach.
From Anqing to Ithaca
Xie grew up in Anqing, a small town in Anhui Province, China, and received her undergraduate degree in mathematics from Fudan University in Shanghai. She says she came to Cornell because of the high rank of ORIE’s program and the attraction of Cornell as a university. Like Ithaca, Anqing is a small, quiet college town, although without a large university and with somewhat less snow in winter, according to Xie.
Xie is the first recipient of funds from a Ph.D. Fellowship Fund endowed by Robert S. Kaplan Ph.D. ‘68, who is now Marvin Bower Professor of Leadership Development, Emeritus at the Harvard Business School. She enjoys music, in particular playing the violin and singing karaoke, and travelling, having taken trips to U.S. National Parks, Alaska, Puerto Rico and other locations since coming to the U.S. After completing her degree, she hopes to work on risk analysis at a financial services firm.