INFORMS 2021 Program Book

INFORMS Anaheim 2021

WB24

face a difficult optimization in determining their reaction and may therefore not pursue a globally optimal solution to their problem. In such cases, the leader seeks a solution that is robust against a wider range of possible actions by the follower. We discuss modeling and solution procedures for such bounded rationality problems using a branch-and-cut framework. 3 - On the Tightness and Scalability of the Lagrangian Dual Bound for the Alternating Current Optimal Power Flow Problem Weiqi Zhang, University of Wisconsin-Madison, Madison, WI, United States, Kibaek Kim, Victor Zavala We study tightness and scalability properties of a Lagrangian dual (LD) bound for the nonconvex alternating current optimal power flow (ACOPF) problem. We show that the LD bound is as tight as that provided by the powerful and popular semidefinite programming relaxation. However, a key advantage of the proposed bound is that it can be computed in a parallel, decentralized manner. Specifically, in the proposed approach we partition the network into a set of subnetworks, we dualize the coupling constraints (giving the LD function), and we maximize the LD function with respect to the dual variables of the coupling constraints (giving the desired LD bound). The dual variables that maximize the LD are obtained by using a bundle method and we provide a proof of convergence for such method.We demonstrate our developments using PGLib test instances. WB27 CC Room 206B In Person: Optimization in Quantum Computing and Vice Versa I General Session Chair: Baoyu Zhou, Lehigh University, Bethlehem, PA, 18015, United States 1 - Automated Design of Magnetic Resonance Pulse Sequences Using Physics-inspired Optimization Stephen Jordan, Microsoft, Redmond, WA, United States Quantum annealing is a form of quantum computation that can solve non- convex optimization problems using tunneling effects to escape from local minima. However, in many cases the benefit of such tunneling phenomena can be replicated by Monte Carlo methods on standard classical computers. Here we describe the application of physics-inspired optimization methods to the automated design of pulse sequences for magnetic resonance imaging. This is a non-convex optimization problem with thousands of variables. Our global optimization methods, starting from random initial states, have produced novel pulse sequences which reduce the duration of brain scans by up to 4x relative to state of the art sequences designed by human experts, while also achieving robustness against systematic error due to magnetic field inhomogeneities in the scanner. 2 - Fast Quantum State Tomography via Accelerated Non-convex Programming J. Lyle Kim, Rice University, Houston, TX, United States We propose a new quantum state reconstruction method, called Momentum- Inspired Factored Gradient Descent (MiFGD), that combines ideas from compressed sensing, non-convex optimization, and acceleration methods. Despite being a non-convex method, MiFGD converges provably to the true density matrix at a linear rate under common assumptions. With this manuscript, we present the method, prove its convergence property and provide Frobenius norm bound guarantees with respect to the true density matrix. From a practical point of view, we benchmark the algorithm performance with respect to other existing methods, in both synthetic and real experiments performed on an IBM’s quantum processing unit. We find that the proposed algorithm performs orders of magnitude faster than state of the art approaches, with the same or better accuracy. 3 - Fast Approximation for Power System State Estimation under Cyber Attacks Kamal Basulaiman, University of Pittsburgh, Pittsburgh, PA, United States Power system state estimation (PSSE) is yet a critical problem. With the increasing emergence of renewable energy resources, the power system states are becoming less predictable. Therefore, there is an urgent need for frontier models that can accurately estimates the system states with the associated surges to monitor the electric power system in an economic and secure fashion. This work addresses the PSSE problem when the data is corrupted by cyber attacks. This problem is non- convex and known to be NP-hard in general. We propose a fast approximation via deep learning that is suitable for real-time setting. Experimental results demonstrate the capability of our model compared to the state-of-the-art.

WB24 CC Room 205A In Person: Technology-driven Emerging Issues in Supply Chains General Session Chair: Junghee Lee, University of Notre Dame, New Orleans, LA, 70118-5669, United States 1 - Supplier Development in a Coopetition Setting Xinxue Qu, University of Notre Dame, Granger, IN, 46530-8209, United States, Hemant K. Bhargava, Daewon Sun Channel partnership can take many forms, including those where partners also engage in direct competition. Manufacturers and retailers partner to sell goods but retailers can often have their own competing products. Similarly, component specialists partner with producers in offering complete goods to consumers, but component makers sometimes also compete with an end-user product. For instance, Samsung supplies OLED screens that are used in Apple’s iPhone X, but Samsung also makes high-end phones that directly compete with Apple’s iPhone X. This study aims to analyze the economic tensions and interplay involved in R&D investment in such market settings. The effect of several market variables, including the intensity of competition in the end-user market, firms’ market advantage, and the efficiency for product development, is also examined. 2 - Supply Chain Relationship Impacts on Firms’ Environmental CSR Practice Marcus A. Bellamy, Boston University, Rafik B. Hariri Building, Boston, MA, 02215, United States, Elliot Bendoly, Erin McKie Our study emphasizes how a firm’s corporate social responsibility (CSR) efforts are linked to their supply chain entities. Specifically, we examine how changes in Environmental CSR occur across members of their supply chains using supply chain relationship and environmental data over a 10-year period. Based on these findings, we discuss the implications of firms’ supply chain engagements and selections. 3 - First or Second Dose first? Two-dose Vaccine Allocation under Finite Capacity Yun Zhou, McMaster University, Hamilton, ON, L9A 0A3, Canada, Ming Hu Most Covid-19 vaccines require two doses for a recipient to be considered as fully vaccinated. However, receiving only one dose still provides some protection. We study the allocation of vaccination capacity between 1st and 2nd doses. Building on the SIR model, we formulate an optimal control model and show that the optimal policy makes an all-or-nothing-type allocation at any time instant. We then compare the “first doses first” and “second doses first” policies. We find that even if the 1-dose efficacy is slightly greater than 50% of the 2-dose efficacy, “second doses first” is still better. WB25 CC Room 205B In Person: Algorithms and Software for Optimization under Uncertainty General Session Chair: Weiqi Zhang, Middleton, WI, 53562, United States 1 - ROC++: Robust Optimization in C++ Qing Jin, University of Southern California, Los Angeles, CA, United States, Phebe Vayanos We propose ROC++, an open source C++ based platform for automatic robust optimization, applicable to a wide array of single- and multi-stage robust problems with both exogenous and endogenous uncertain parameters, that is easy to both use and extend. It also applies to certain classes of stochastic programs involving continuously distributed uncertain parameters and decision- dependent information discovery. We also offers ROPy, a Python interface in the form of a callable library. We showcase the modeling power of ROC++ on several decision-making problems of practical interest. Our platform can help streamline the modeling and solution of stochastic and robust optimization problems for both researchers and practitioners. 2 - Mixed Integer Bilevel Linear Optimization with Bounded Rationality Yu Xie, Lehigh University, Bethlehem, PA, United States, Ted K. Ralphs We consider mixed integer linear bilevel optimization problems in which the leader makes an initial decision, and then the follower reacts. Unlike the traditional setting, where decision-makers are assumed to have complete information and act rationally, we consider situations in which the follower may

154