Informs Annual Meeting 2017

MC83

INFORMS Houston – 2017

Monday, 3:10 - 4:00PM

3 - The Stochastic Container Relocation Problem: Sampling in Multi-stage Stochastic Optimization with Bounded Cost Virgile Galle, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Building E40-130, Cambridge, MA, 02139, United States, vgalle@mit.edu, Setareh Borjian Boroujeni, Vahideh Manshadi, Cynthia Barnhart, Patrick Jaillet The Container Relocation Problem (CRP) is concerned with finding a sequence of moves of containers that minimizes the number of relocations needed to retrieve all containers, while respecting a given order of retrieval. This paper studies the stochastic CRP, which relaxes the assumption of knowing the full retrieval order of containers (particularly unrealistic in real operations). We introduce a new multi-stage stochastic model, and propose an optimal algorithm based on pruning as well as a near-optimal algorithm with bounded average error. Algorithms are tested on existing instances. 382C Stochastic Optimization and Variational Inequality Problems Sponsored: Optimization, Optimization Under Uncertainty Sponsored Session Chair: Konstantin Pavlikov, University of Florida, 1350 N. Poquito Road, Shalimar, FL, 32579-1163, United States, kpavlikov@ufl.edu 1 - SI-ADMM: A Stochastic Inexact ADMM Framework for Resolving Structured Stochastic Convex Programs We consider the resolution of the structured stochastic convex program: min E[f(x, )] + E[g(y, )] such that Ax + By = b. To exploit problem structure and allow for developing distributed schemes, we propose a stochastic inexact ADMM (SI- ADMM) in which the subproblems are solved inexactly via stochastic approximation schemes. Based on this framework, we prove that (i) when the inexactness sequence satisfies suitable summability properties, SI-ADMM produces a sequence that converges to the unique solution a.s.; (ii) if the inexactness drops at a geometric rate, the sequence converges to the unique solution in a mean-squared sense at geometric rate, and the total sample complexity can be canonical. 2 - On the Variants of Projection and Extragradient Schemes for Stochastic Variational Inequality Problems Shisheng Cui, Pennsylvania State University, University Park, PA, 16801, United States, cuishisheng05@gmail.com, Shanbhag Uday The stochastic generalizations of the extragradient methods are complicated by a key challenge: the scheme requires two projections on a convex set and two evaluations of the map for every major iteration. It becomes a concern if the sets are complicated. Thus, we consider two related avenues where every iteration requires a single projection: (i) A projected reflected gradient method; and (ii) A subgradient extragradient method. We prove the almost sure convergence of the iterates to a random point in the solution set and show that the gap function associated with the averaged sequence diminishes to zero at the optimal rate. 3 - Multi-period Probabilistic Set Covering Problem Konstantin Pavlikov, University of Florida, 1350 N. Poquito Road, Shalimar, FL, 32579-1163, United States, kpavlikov@ufl.edu This study extends the classical probabilistic set coverage problem to a multi- period setting. The problem focuses on finding a least cost facility locations to cover random demand over one period of time with a specified probability. If facility engagement lasts more than one period of time, then little is known about how well the remaining operational facilities can satisfy demand in period two. Hence, the two-period model maximizes the probability of satisfying demand in period two by facilities not engaged in satisfying demand from period one. We demonstrate how solving multi-period version of the problem can be reduced to solution of a sequence of one-period probabilistic set coverage problems. MC83 Yue Xie, Pennsylvania State University, 445 Waupelani Dr, Pennsylvania, PA, 16801, United States, yux111@psu.edu, Uday Shanbhag

Keynote General Assembly A, Level 3 Keynote- INFORMing Process Improvement and Patient Safety in Healthcare Invited Session 1 - INFORMing Process Improvement and Patient Safety in Healthcare Victoria Jordan, Emory Because of my previous position as Executive Director, Performance Improvement at the University of Texas (UT) M. D. Anderson Cancer Center, the conference committee asked me to share information about and the history behind the Texas Medical Center and MD Anderson. I will also discuss my role at UT as Chancellor’s Fellow for Systems Engineering and how our team was able to establish partnerships between university faculty (in Business and Engineering) and healthcare professionals to spread the use of operations research and management science in healthcare. I will share some exciting accomplishments from those efforts as well as more general applications across the healthcare industry. I hope you will leave this session with a new understanding of the challenges in healthcare and an interest in using Operations Research and Management Science to address some of these challenges. Keynote General Assembly B, Level 3 Keynote: Systems Approach to Managing Risk in Human Spaceflight Missions Invited Session 1 - Systems Approach to Managing Risk in Human Spaceflight Missions Nancy Currie-Gregg, Texas A&M.(formerly NASA) Human spaceflight is an inherently risky endeavor. From the first human space exploration missions to recent problems on the International Space Station, NASA has faced many challenges and has relied on creative and innovative engineering solutions to complete the mission and safely return the crew. However, over the past 50 years, NASA has also experienced three fatal accidents resulting in the loss of seventeen crewmembers. Addressing the complex interdependencies and systemic causes associated with these tragedies, including those associated with organizational culture and management, provides insight to the advantages of a systems approach to managing risk in high reliability organizations. Keynote Grand Ballroom C, Level 3 Keynote: IFORS Distinguished Lecture Invited Session 1 - Biased Random-Key Genetic Algorithms: Components, Evolutionary Dynamics and Applications Celso Ribeiro, Universidade Federal Fluminense A biased random-key genetic algorithm (BRKGA) is a general search procedure for finding optimal or near-optimal solutions to hard combinatorial optimization problems. It is derived from the random-key genetic algorithm of Bean (1994), differing in the way solutions are combined to produce offspring. BRKGAs have three key features that specialize genetic algorithms. First, a fixed chromosome encoding using a vector of N random keys over the real interval [0, 1), where the value of N depends on the instance of the optimization problem. Second, a well- defined evolutionary process adopting a parameterized uniform crossover to generate offspring and thus evolve the population. Third, the introduction of new chromosomes called mutants in place of the mutation operator usually found in evolutionary algorithms. Such features simplify and standardize the procedure with a set of self-contained tasks from which only one is problem-dependent: that of decoding a chromosome, i.e. using the keys to construct a solution to the underlying optimization problem, from which the objective function value or fitness can be computed. In this talk, we review the basic components of a BRKGA and introduce a framework for quick implementations of BRKGA heuristics. We then illustrate the application of this framework to a few case studies in a network routing, load scheduling and data mining problems. We conclude with a brief review of other domains where BRKGAs have been applied.

234