Clark School Home UMD

ISR News Story

A new unbiased stochastic derivative estimator for simulating complex systems

Professor Michael Fu (BMGT/ISR) and his international colleagues have recently published research in the journal Operations Research, A new unbiased stochastic derivative estimator for discontinuous sample performances with structural parameters.” This paper proposes a new unbiased stochastic derivative estimator in a framework that can handle discontinuous sample performances with structural parameters. Fu's co-authors are Peking University’s Yijie Peng, Fudan University’s Jian-Qiang Hu, and Vrije Universiteit Amsterdam’s Bernd Heidergott.

With the permission of the University of Maryland’s Robert H. Smith School of Business, we reprint below a very interesting article about this research written by Karen Johnson, a senior communications writer at the Smith School. Many thanks to our Smith School colleagues!

Estimating Sensitivities in the Simulation of Complex Systems

by Karen Johnson

Numbers can be a tricky business.

Michael C. Fu, the Smith Chair of Management Science and chair of the Decision, Operations and Information Technologies Department at the University of Maryland’s Robert H. Smith School of Business, understands just how tricky they can be.

Fu, who has a joint appointment with the Institute for Systems Research and an affiliate appointment with the Department of Electrical and Computer Engineering, both in the A. James Clark School of Engineering, recently co-authored a research article with Peking University’s Yijie Peng, Fudan University’s Jian-Qiang Hu, and Vrije Universiteit Amsterdam’s Bernd Heidergott, in which he examined some of the difficulties in getting the numbers right when simulating complex systems.

The paper, “A New Unbiased Stochastic Derivative Estimator for Discontinuous Sample Performances With Structural Parameters,” provides an efficient and accurate methodology for estimating sensitivities in complex systems that require simulation for performance evaluation, specifically, by introducing a new method that can handle discontinuous sample performance measures — an open research problem with various practical applications.

The estimator the researchers formulated is unbiased and has an analytical form, and the general framework addresses many derivative estimation problems, for example in settings of pricing financial derivatives, analyzing queueing systems, performing inventory control and management, doing statistical process control, or managing supply chains, manufacturing plants, transportation systems, communications networks, or service systems such as banks, amusement parks, retail stores.

Fu offers a quote from a book popular with practitioners in the financial services industry, “Monte Carlo Methods in Financial Engineering,” which says, “Whereas the prices themselves can often be observed in the market, their sensitivities cannot, so accurate calculation of sensitivities is arguably even more important than calculation of prices.” As examples, Fu says, consider how a finance professional might estimate the delta of an IBM stock option as compared to determining the price of the option itself.

An important statistical property of the estimator is that it be unbiased, the researchers explain, meaning that the expected value of the estimator equals the quantity that it is trying to estimate (for example, the price sensitivity of a stock option). The new methodology can handle both continuous and discontinuous settings. Continuous performance measures have been addressed by past research, whereas discontinuous performance measures are more challenging. Binary outcomes are the simplest examples of discontinuous performance measures. “In the context of finance, a digital option pays off a fixed value (e.g., $100) if above a certain threshold —the strike price — and zero otherwise,” Fu explains.

Instances of a structural parameter in the stock option example would be the strike price or expiration date, in contrast to a distributional parameter, which would affect the dynamics of the underlying asset.

| View this article on the Smith School’s website |

Related Articles:
Michael Fu wins NSF grant for simulation-based optimal decision making
Kedem, De Oliveira win Canadian Journal of Statistics Award
Researchers write about Google's AlphaGo for OR/MS Today
Maryland research contributes to Google’s AlphaGo AI system
Alumna Enlu Zhou earns tenure at Georgia Tech
Marcus, Fu receive NSF grant for optimization research
Ph.D. student Kevin Lister wins Link Foundation Fellowship
Alum Amnon Lotem promoted to Chief Technology Officer at Skybox Security
Work by Zhou, Lin, Marcus and Fu wins 'Best Theoretical Paper' award
Marcus, Fu receive NSF grant for particle filtering for stochastic control and global optimization

March 22, 2018

Prev   Next



Current Headlines

ECE Names 2017-2018 Distinguished Dissertation Fellows

UMD Celebrates Invention of Year Winners, Ventures, and Partnerships at 'Innovate Maryland'

Ulukus named Anthony Ephremides Professor in Information Sciences and Systems

Maryland researchers develop computational approach to understanding brain dynamics

Alumnus Raef Bassily joins Ohio State as tenure-track faculty

Maryland researchers awarded $1M DARPA Lagrange program cooperative agreement

Bill Regli featured at Global Industrial Cooperation Conference

Brin Family Prize to Support Excellence in Drone-Related Activity

Schizophrenia drug monitoring device research featured on IEEE Sensors Letters cover

Maryland researchers awarded DARPA cooperative agreement to develop robotic swarm strategies

News Resources

Return to Newsroom

Search News

Archived News

Events Resources

Events Calendar