Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TE01

Tuesday, 4:35PM - 6:05PM

parameters is often not possible in practice. This study addresses MDPs under cost and transition probability uncertainty with the aim of obtaining policies which optimize the Value-at-Risk associated with the expected performance of an MDP model in terms of parameter uncertainty. Considering a sampling approach, we provide a mixed-integer programming formulation and a branch-and-cut algorithm. Our proposed methods are demonstrated on an inventory management problem for humanitarian relief operations during a slow onset disaster. 2 - Two-stage Distributionally Robust P-order Conic Mixed Integer Programs Yingqiu Zhang, Virginia Tech, 305 Loudon Rd #631, Blacksburg, VA, 24060, United States We present two-stage distributionally robust p-order conic integer programs (TSDR-CMIPs) in which the second-stage problems have p-order conic constraints along with integer variables. We provide sufficient conditions under which the integrality constraints on the second-stage integer variables can be relaxed, without effecting the integrality of the optimal solution, by adding parametric (non)-linear inequalities. We introduce structured CMIPs in the second stage of TSDR-CMIPs, derive inequalities which satisfy the foregoing conditions, and present results of our extensive computational experiments for p=2. 3 - SIPLIB 2.0: Stochastic Integer Programming Library 2.0 Yongkyu Cho, Pohang University of Science and Technology, Pohang, Korea, Republic of, Kibaek Kim, James Luedtke, Jeff T. Linderoth SIPLIB was constructed in 2002 by Shabbir Ahmed and his colleagues. The library has been providing a collection of test instances to facilitate computational and algorithmic research in SIP. State-of-the-art in SIP combined with the speedup in computing machinery, however, has made many SIPLIB instances trivial. By SIPLIB 2.0, we provide 1) richer collection of test instances with benchmarking computational results, 2) not only SMPS files but also utilities for generating/analyzing instances. SIPLIB 2.0 is implemented in Julia programming language with modeling package StructJuMP (block-structured optimization framework for JuMP). 4 - Generalized Alpha-approximations for Mixed-integer Recourse Models Ward Romeijnders, University of Groningen, Groningen, Netherlands, Niels van der Laan We consider generalized alpha-approximations for two-stage mixed-integer recourse models. They are a generalization of a class of convex approximations developed for simple integer recourse models and totally unimodular integer recourse models. We derive an error bound for these generalized alpha- approximations that converges to zero if the total variations of the probability density functions of the random variables in the model converge to zero. n TE03 North Bldg 121C Innovative Business Models and Bottom of the Pyramid Sponsored: Manufacturing & Service Oper Mgmt Sponsored Session Chair: Serguei Netessine, The Wharton School, Philadelphia, PA, 19104, United States Co-Chair: Bhavani Shanker Uppari, INSEAD, Singapore, 138676, Singapore 1 - Disruptions, Resilience and Performance of Emerging Market Entrepreneurs: Evidence from Uganda Amrita Kundu, PhD Candidate, London Business School, London, United Kingdom, Stephen J. Anderson, Kamalini Ramdas We examine the effect of firm-specific business disruptions (both managerial and operational) on the performance of small firms in emerging markets and the effectiveness of appropriate resilience strategies in buffering against these disruptions. Using a hand-built panel dataset on 646 small firms over four time periods in Kampala, Uganda, we find that disruptions are highly prevalent and have a statistically and economically significant effect on the performance of these firms. Importantly, we find that both relational and resource resilience significantly help buffer against the negative impact of managerial and operational disruptions, respectively.

n TE01 North Bldg 121A Distributionally Robust Optimization Sponsored: Optimization/Optimization Under Uncertainty Sponsored Session Chair: Grani Adiwena Hanasusanto, The University of Texas at Austin, Austin, TX, 78712, United States 1 - A Data-driven Distributionally Robust Optimization Approach for Appointment Scheduling with Random Service Durations and No-shows Guanglin Xu, University of Minnesota, Minneapolis, MN, 55108, United States, Ruiwei Jiang We study a single-server appointment-scheduling problem where the number of appointees and the sequence of their arrivals are fixed. The service durations and appointees’ show-ups are stochastic, but the true probability distribution is unknown. With a collection of historical data, we propose a data-driven distributioanlly robust optimization model, which minimizes the worst-case total expected cost, by optimizing the scheduled arrival times of the appointees. We reformulate the resulting problem into a copositive program and discuss tractabilities under mild conditions. We then develop a tight semidefinite- programming-based approximation and validate it on benchmark instances. 2 - Distributionally Robust Optimization with Infinitely Constrained Ambiguity Sets Melvyn Sim, National University of Singapore, 15 Kent Ridge Building, Biz1 8-76, Singapore, 119245, Singapore, Huan Xu, Zhi Chen We propose a new class of infinitely constrained ambiguity sets for which the number of expectation constraints could be infinite. The description of such ambiguity sets can incorporate the stochastic dominance, the dispersion, the fourth moment, and our newly proposed “entropic dominance information about the uncertainty. 3 - A Copositive Approach for Multi-stage Robust Optimization Problems Grani A. Hanasusanto, The University of Texas at Austin, Austin, TX, 78712, United States, Guanglin Xu In this talk, we study generic multi-stage linear robust optimization problems. We employ affine decision rules for problems with both uncertain objective coefficients and recourse matrices, and utilize quadratic decision rules for problems with only uncertain right-hand sides. The emerging optimization problems are NP-hard but amenable to copositive programming reformulations that give rise to tight conservative approximations. We provide both theoretical and numerical results to demonstrate the effectiveness of the copositive programming approach. 4 - On the Heavy-tail Behavior of the Distributionally Robust Newsvendor Model Karthik Natarajan, Singapore University of Technology and Design, Singapore University of Technology and Design, 20 Dover Drive, Singapore, 138682, Singapore, Bikramjit Das, Anulekha Dhara Since the seminal work of Scarf (1958) on the newsvendor problem with ambiguity in the demand distribution, there has been a growing interest in the study of the distributionally robust newsvendor problem. A simple observation indicates that the optimal order quantity in Scarf’s model for any possible value of the critical ratio is also optimal for a censored student-t distribution with degrees of freedom parameter 2 that has infinite variance. In this paper, we generalize this heavy-tail optimality property to the case when information on the first and the nth moment is known for any real number n > 1. n TE02 North Bldg 121B Two-Stage Stochastic Mixed-Integer Programming Sponsored: Optimization/Optimization Under Uncertainty Sponsored Session Chair: Yingqiu Zhang, Virginia Tech, Blacksburg, VA, 24060, United States 1 - Markov Decision Processes Under Parameter Uncertainty with a Chance-constrained Programming Approach Merve Merakli, University of Washington, Simge Kucukyavuz In classical Markov Decision Processes (MDPs), action costs and transition probabilities are assumed to be known, although an accurate estimation of these

380