Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SD38

and study the impact of such information on the achievable performance and the design of efficient decision-making policies. We characterize the regret complexity of this family of problems and introduce a general and practical adaptive exploration approach for designing policies that, without any knowledge on the information arrival process, attain the best performance that is achievable when that arrival process is a priori known. 2 - Stochastic Game for Fuel Follower Problem: N vs MFG Renyuan Xu, University of California-Berkeley, 4141 Etcheverry Hall, OR Department, Berkeley, CA, 94720, United States In general, finding a Nash equilibrium (NE) or proving Pareto optimality (PO) for continuous-time stochastic game with multiple players is hard, even when there are only two players in the game. Mean Field Game (MFG) provides a powerful tool and analytically feasible framework to approximate the NE and the PO of the N-player stochastic games. In this talk, we will first define the stochastic game for the classical fuel follower problem in the seminal paper by Benes, Shepp, and Witsenhausen (1980). We will then solve and compare with explicit solutions for the NE and the PO of the N-player game, and the corresponding MFG when the number of players goes to infinity. This is a joint work with Xin Guo (UC Berkeley). 3 - Statistics of Robust Optimization: A Generalized Empirical Likelihood Approach Hongseok Namkoong, Stanford University, Stanford, CA, 94305, United States We study statistical inference and distributionally robust solution methods for stochastic optimization problems, focusing on confidence intervals for optimal values and solutions that achieve exact coverage. We develop a generalized empirical likelihood framework—-based on distributional uncertainty sets constructed from nonparametric f-divergence balls—-for Hadamard differentiable functionals, and in particular, stochastic optimization problems. Our theory prescribes the robustness level that provide one- and two-sided confidence intervals that achieve exact coverage. 4 - SOAP: One Clean Analysis of All Age-Based Scheduling Policies Ziv Scully, Carnegie Mellon University, 5000 Forbes Avenue, GHC 7121, Pittsburgh, PA, 15213, United States We consider an extremely broad class of M/G/1 scheduling policies called SOAP: Schedule Ordered by Age-based Priority. The SOAP policies include most policies in the literature as well as infinitely many variants which have never been analyzed. SOAP policies range from classic policies, like first-come, first-serve (FCFS), foreground-background (FB), class-based priority, and shortest remaining processing time (SRPT); to much more complicated scheduling rules, such as the famously complex Gittins index policy and shortest expected remaining processing time (SERPT). We present a universal analysis of all SOAP policies, deriving the mean and Laplace-Stieltjes transform of response time. n SD38 North Bldg 225B Stochastic Models for the Blockchain Sponsored: Applied Probability Sponsored Session Chair: Ciamac Cyrus Moallemi, Columbia University, New York, NY, 10027, United States Co-Chair: Jacob Leshno, Columbia University, New York, NY, 10027, United States 1 - Bitcoin: A Natural Oligopoly Nicholas A. Arnosti, Columbia Business School, 3022 Broadway, Uris Hall rm 402, New York, NY, 10027, United States Although Bitcoin was intended to be a decentralized digital currency, in practice, mining power is quite concentrated. This fact is a persistent source of concern for the Bitcoin community. We provide an explanation using a simple model to capture miners’ incentives to invest in equipment. In our model, n miners compete for a prize of fixed size. Each miner chooses an investment level, incurring cost proportional to their investment. Miners receive rewards proportional to their investment. We show that seemingly small cost differences result in highly concentrated mining power, and discuss the implications of our results for cryptocurrency design. 2 - Scaling Cryptocurrency Blockchains Rhys Bowden, Rhys Bowden, Anthony Krzesinski, Peter G. Taylor Cryptocurrency blockchains form the globally distributed ledgers of their associated cryptocurrencies. A blockchain is updated at random intervals, when a miner adds a new block (referencing the most recent block they have heard about) on to the end of the chain, then that block is communicated to the other miners and the rest of the network as a whole. Typically the time taken to communicate a new block is much less than the average time between blocks arriving. We model this process to show how the system functions when blocks arrive so frequently that consensus is rare.

n SD35 North Bldg 224A

UAS and Aviation Forecast Sponsored: Aviation Applications Sponsored Session Chair: Chia-Mei Liu, Federal Aviation Administration, Washington, DC, 20591, United States 1 - Airline Network Evolution Forecasting Reed Harder, Dartmouth, 14 Engineering Drive, Hanover, NH, 03755, United States, Vikrant Vaze Airline networks evolve as airlines enter and exit various markets and routes. Accurate modeling of this evolution would allow for better forecasting of airport activity and more effective long-term planning by decision makers in the air transportation system. However, forecasting airline network evolution remains a challenge because of the complicated, interdependent nature of available data and the relative sparsity of entry and exit events. We present approaches for the probabilistic modeling of airline network evolution that leverage multiple sources of information on potential market and route entries. 2 - Unmanned Aircraft System (UAS) National Forecast in the U.S. Bhadra Dipasis, Federal Aviation Administration Abstract not available. 3 - Air Route Traffic Control Center (ARTCC) Operation Forecast David Hechtman, Senior Modeling and Simulation Engineer, The MITRE Corporation, 7515 Colshire Drive, McLean, VA, 22102- 7539, United States, Michael Yablonski, Ricky Marske The FAA Office of Aviation Policy and Plans requested that The MITRE Corporation develop a forecasting method for Air Route Traffic Control Center (ARTCC) operations using an origin-and-destination (O-D) passenger demand framework. FAA’s Terminal Area Forecast-Modernization (TAF-M) predicts O-D segment demand but not ARTCC operations. To provide this information, MITRE developed a TAF-based ARTCC forecast incorporating O-D segment demand from TAF-M and historical routing data from the FAA Traffic Flow Management System. 4 - The Effect of Oil Price Pass Through Rate on U.S. Domestic Airfare Forecast Chia-Mei Liu, Federal Aviation Administration, 800 Independence Avenue SW, Washington, DC, 20591, United States, Kang Hua Cao FAA publishes Terminal Area Forecast (TAF) annually. TAF includes enplanement and operation forecasts for all U.S. airports. Among the heavily traveled airports, FAA also publishes optimistic and pessimistic forecasts to quantify forecast uncertainty. One important driver for the forecast uncertainty is the uncertainty in oil price, which leads to the uncertainty in airfares and enplanements subsequently. This research studies airlines’ ability to pass oil price change to the consumers by raising or reducing airfares. We control for factors such as airline competition and economic activities to quantify the airfare elasticity with respect to oil price change. n SD36 North Bldg 224B Joint Session AAS/Practice Curated: Aviation Applications Section: Awards Finalists Sponsored: Aviation Applications Sponsored Session Chair: Vikrant Vaze, Dartmouth College, 14 Engineering Drive, n SD37 North Bldg 225A APS Student Paper Competition Sponsored: Applied Probability Sponsored Session Chair: John Hasenbein, University of Texas-Austin, 1 University Station Stop C2200, Department of Mechanical Engineering, Austin, TX, 78712-0292, United States 1 - Adaptive Learning with Unknown Information Flows Ahmadreza Momeni, Stanford University, Stanford, CA, 94025, United States We introduce a generalized multi-armed bandit formulation in which additional information on each arm may appear arbitrarily throughout the decision horizon, Murdough Center, Hanover, NH, 03755, United States 1 -Aviation Applications Section: Awards Finalists

103