Air Force Office of Scientific Research Designates Peter Frazier an Outstanding Young Investigator
ORIE Assistant Professor Peter Frazier is one of 43 scientists and engineers throughout the United States to submit a winning research proposal through the Air Force's Young Investigator Research Program. Frazier, who joined the ORIE faculty in 2009, proposes to find new algorithms for calibrating and optimizing simulations of systems occurring in nature, engineering, combat, and individual and collective human behavior. He proposes to explore "Decision-Theoretic Methods in Simulation Optimization."
Nature of Simulation Optimization
Typically, simulation techniques are employed when it is difficult or impossible to represent the system with equations that can then be solved to predict and achieve the best outcomes. Among other sources of difficulty is inherent randomness in the system, which leads to outcomes expressed in probabilistic terms. While simulation techniques can take this randomness into account, many simulation runs are required just to predict the outcome of each possible decision, much less to find the best decision. Hence Frazier faces a particular challenge in building algorithms that find the optimum in a system that is represented by a simulation rather than equations.
Frazier sees simulation optimization as a decision-theory problem in sequential experimental design, with the objective of finding solutions with minimal error and computational effort. He has developed several promising conjectures based on this approach. "Proofs of these conjectures would constitute an important theoretical advance, and would provide new high-performance algorithms," he points out.
Frazier can be seen in this CornellCast interview, where he explains a simple and commonplace example of the kind of problem to which his theoretical work has application: how much to turn the faucet in a shower to achieve the right temperature. He explains that the problems discussed in his winning proposal "are essentially this problem, except that instead of one faucet we have 5 or 10 or 20 faucets, and instead of waiting a couple of hours or perhaps even days for the result. And instead of finding the right temperature in the shower, the goal is to calibrate a model of climate change, accurately reconstruct a collection of whole genomes from fragmented genetic data, or find optimal staffing levels and schedules for a large hospital."
A Broad Background
A graduate in Physics and Engineering & Applied Sciences from the Calfornia Institute of Technology, Frazier received his Ph.D. in Operations Research and Financial Engineering from Princeton. He worked as a software engineer prior to pursuing graduate education. His research applies an approach known as fully sequential Bayesian analysis to a broad array of theoretical and practical problems he characterizes as 'optimal learning.'
Frazier is one of three Cornell faculty members among the 43 researchers nationwide to be selected by the Air Force as outstanding young investigators. Assistant Professor of Physics Kyle Shen received an award to analyze a spectroscopic search for new superconductors in artificial materials. Assistant Professor of Electrical and Computer Engineering Salman Avestimehr received an award for research related to the design of distributed communication networks for military applications. A total of $16.5 million in grants have been awarded in the program this year.
For additional information, click here.