INFORMS 2021 Program Book

INFORMS Anaheim 2021

MD10

Chair: Bistra Dilkina, University of Southern California, Los Angeles, CA, 90089, United States 1 - Optimal Land Supply for BECCS Considering Biodiversity Conservation Cindy Azuero, Georgia Tech, Atlanta, GA, United States Bioenergy with carbon capture and storage (BECCS) will play a major role in mitigation pathways toward the 1.5º and 2ºC scenarios. Estimated land requirements for BECCS are big, ranging from 200 Mha-1500 Mha, for a deployment between 3-30 Gt CO2 per year in 2100 (Creutzig et al., 2021). Current land allocation models used in Integrated Assessment Models (IAMs) do not consider biodiversity impacts when determining the location of the land supplied for bioenergy. Here, we integrate a linear optimization model with a biodiversity impact assessment model, to determine how to optimally supply land for BECCS considering (1) minimizing biodiversity impact subject to a budget constraint and (2) minimizing cost subject to a biodiversity threshold. A pareto frontier is constructed with which the trade-off between biodiversity and cost can be analyzed. 2 - Interdiction of Wildlife Trafficking Supply Chains: An Analytical Approach Burcu B. Keskin, University of Alabama, Tuscaloosa, AL, 35406- 4062, United States, Bistra Dilkina, Aaron Ferber, Emily Barbee, Oakley Prell Illicit Wildlife Trafficking (IWT) has a negative impact on the environment and communities, enabling the spread of diseases, land degradation, and biodiversity loss. IWT is a global issue, with almost 6,000 different species seized since the 90s and expanding to more than 150 countries. Traffickers operate complex and dynamic networks that require detailed strategies to disrupt. We model disrupting these networks as a network interdiction problem where authorities seek to interdict along specific routes to reduce the trafficker’s profitability and force any resulting geographical displacement to be as costly as possible. We characterize the needed and available data in IWT, modeling assumptions, and network interdiction formulations that apply to this complex setting, and we evaluate the proposed methods in the context of global air travel networks. MD13 CC Room 201A In Person: Stochastic First-order Methods for Constrained Optimization General Session Chair: Afrooz Jalilzadeh, The University of Arizona, Tucson, AZ, United States 1 - Inexact-proximal Accelerated Gradient Method for Stochastic Nonconvex Constrained Optimization Problems Morteza Boroun, University of Arizona, Tucson, AZ, United States, Afrooz Jalilzadeh Stochastic nonconvex optimization problems with nonlinear constraints have a broad range of applications in intelligent transportation, cyber-security, and smart grids. In this paper, first, we propose an inexact-proximal accelerated gradient method to solve a nonconvex stochastic composite optimization problem where the objective is the sum of smooth and nonsmooth functions and the solution to the proximal map of the nonsmooth part is calculated inexactly at each iteration. We demonstrate an asymptotic sublinear rate of convergence for stochastic settings using increasing sample-size considering the error in the proximal operator diminishes at an appropriate rate. Then we customize the proposed method for solving stochastic nonconvex optimization problems with nonlinear constraints and demonstrate a convergence rate guarantee. 2 - A Stochastic Variance-reduced Accelerated Primal-dualmethod For Finite-sum Saddle-point Problems Erfan Yazdandoost Hamedani, University of Arizona, State College, PA, 16801-4415, United States, Afrooz Jalilzadeh In this talk, we propose a variance-reduced primal-dual algorithm for solving convex-concave saddle-point problems with finite-sum structure and nonbilinear coupling function. This type of problem typically arises in machine learning and game theory. Compared with existing methods, our framework yields a significant improvement over the number of required primal-dual gradient samples to achieve an epsilon-accuracy of the primal-dual gap. We implemented our method for solving a distributionally robust optimization problem to show the effectiveness of the proposed algorithm. 3 - Using Deep Reinforcement Learning for Solving the Stochastic Capacitated Lot Sizing Problem Lotte van Hezewijk, Eindhoven University of Technology, Eindhoven, 3526 WD, Netherlands Lotte van Hezewijk, ORTEC, Zoetermeer, Netherlands, Nico P. Dellaert, Noud Gademann We study a multi-item stochastic capacitated lot sizing problem. Inspired by industrial cases, we consider a limited production capacity, stochastic demand, and setup times. The objective is to determine the production quantities, while

MD10 CC Room 304B In Person: Sequential Learning by Experimentation Joint Session Chair: Daniel Russo, Columbia University, New York, NY, 10027, United States 1 - Online Advertising via Bandit Experiments: An Efficient Method Suitable for High-dimensional Problems Wenjia Ba, Stanford University, Stanford, CA, 94305, United States, Michael Harrison, Harikesh Nair We consider models of sequential decision-making by an online advertiser. In a sequence of trials, the advertiser first chooses the audience segment to purchase an impression, then chooses the ad for display, and finally observes a binary outcome. The problem becomes high-dimensional if there exist many possible combinations of user and ad choices. Adopting the multi-arm bandit framework, we propose and evaluate an approach (PMDL) that is based on a Poisson regression model, using the debiased Lasso method of Javanard and Montenari (2017) to estimate parameters of that model. In numerical experiments, the performance of PMDL is comparable to that of leading alternatives in low- dimensional settings, and it continues to show good performance in high-dimensional as well as real-data settings where existing alternative methods are computationally infeasible. 2 - Adaptivity and Confounding in Multiarmed Bandit Experiments Daniel Russo, Columbia University, New York, NY, 10027, United States Abstract: Bandit algorithms minimize experimentation costs by adapting effort away from poorly performing arms as feedback is observed. But this feature makes them sensitive to confounding. For instance, popular algorithms can’t address the problem of identifying the best action when day-of-week effects may confound inferences. In response, we propose deconfounded Thompson sampling, (DTS) which makes critical modifications to the way Thompson sampling is usually applied. Our results suggest DTS strikes a delicate balance between adaptivity and robustness to confounding. It attains asymptotic lower bounds on the number of samples required to confidently identify the best action —- suggesting optimal adaptivity —- but also satisfies strong performance guarantees in the presence of day-of-week effects and delayed observations —- suggesting robustness. MD11 CC Room 304C In Person: Marketplaces: Empirics and Theory General Session Chair: Thayer Morrill, NC, United States 1 - Preparing for the Worst but Hoping for the Best: Robust (Bayesian) Persuasion Piotr Dworczak, Northwestern University, Evanston, IL, United States, Alessandro Pavan We propose a robust solution concept for Bayesian persuasion that accounts for the Sender’s concern that her Bayesian belief about the environment—which we call the conjecture—may be false. Specifically, the Sender is uncertain about the exogenous sources of information the Receivers may learn from, and about strategy selection. She first identifies all information policies that yield the largest payoff in the “worst-case scenario,” i.e., when Nature provides information and coordinates the Receivers’ play to minimize the Sender’s payoff. Then, she uses the conjecture to pick the optimal policy among the worst-case optimal ones. 2 - Equilibrium Inefficiency in Matching Markets with Interviews Erling Skancke, Stanford University, Stanford, CA, United States Recent debate in the medical literature has brought to light issues with the pre- match interview process for residency positions at hospitals. In this paper, I build a game-theoretic model in which hospitals must simultaneously choose which doctors to interview. Increased interview activity by a hospital always has a negative welfare effect on its competitors, while the strategic externality can be decomposed into two opposing terms. When interview costs are low, hospitals interview more when their competitors do, and the equilibrium exhibits an inefficiently high number of interviews. Moreover, an increase in market size may exacerbate the problem of excessive interviewing.

MD12 CC Room 304D In Person: OR and AI approaches for Biodiversity

Conservation General Session

76