INFORMS 2021 Program Book

INFORMS Anaheim 2021

TB09

hospitalizations, and deaths, incorporating uncertainties in disease transmission as well as impacts of policies, such as masking and facility interventions, and population-level disease prevalence and immunity.

TB04 CC Ballroom D / Virtual Theater 4

Hybrid JFIG Paper competition II Sponsored: Junior Faculty Interest Group Sponsored Session Chair: Manish Bansal, Virginia Tech., Blacksburg, VA, 24061-1019, United States Co-Chair: Dorit Simona Hochbaum, University of California-Berkeley, Berkeley, CA, 94720-1777, United States Co-Chair: Alice E Smith, Auburn University, Auburn, AL, 36849, United States TB05 CC Ballroom E / Virtual Theater 5 Hybrid Getting Started in Public Sector Operations Research Sponsored: Public Sector OR Sponsored Session Chair: Jessica Heier Stamm, Kansas State University, Manhattan, KS, 66503-8514, United States 1 - Getting Started in Public Sector Operations Research I Jessica Heier Stamm, Kansas State University, Manhattan, KS, 66503-8514, United States Public sector operations research is dedicated to decision problems with outcomes that can impact the public / society at large. Such problems often involve complex situations, uncertainty, and multiple stakeholders with differing and potentially conflicting objectives. This panel features speakers with complementary expertise in public sector operations research, who will discuss how to get started in this broad and impactful domain. Panelists will briefly describe their own research, and discussion will proceed to a moderated question and answer session. Questions from the audience are welcomed. 2 - Panelist Austin Buchanan, Oklahoma State University, Stillwater, OK, 74078-5017, United States 3 - Panelist Ozlem Ergun, Northeastern University, Boston, MA, 2115, United States 4 - Panelist Benjamin D. Leibowicz, University of Texas-Austin, Austin, TX, 78712-1591, United States TB06 CC Room 303A In Person: Simulation Applications for Public Policy General Session Chair: Tessa Swanson, Ann Arbor, MI, 48103, United States 1 - A Customizable Agent-based Simulation Tool for Analyzing Infectious Disease Control Strategies in Metropolitan Areas Ashkan Negahban, The Pennsylvania State University, Malvern, PA, United States Non-pharmaceutical interventions such as social distancing, school/business closures, random testing and quarantines are crucial in controlling the spread of an infectious disease. We propose a customizable agent-based simulation and decision support tool that allows for any city to create a fine-grained simulation of the corresponding real-world population and their interactions to enable analysis of various control strategies. We illustrate the applicability and efficacy of the proposed tool for the case of Covid-19 outbreak in New York City. 2 - Simulating Covid-19 Risks Associated with Returning to In-person College Classes Tessa Swanson, Industrial and Operations Engineering, Ann Arbor, MI, United States, Seth Guikema As universities prepare for a school year following disruption from the COVID-19 pandemic, risk analysis can support decision-making for resuming in-person instruction. A simulation-based risk analysis approach enables scenario evaluation and comparison to guide decision making under uncertainty. We develop a simulation model to evaluate various scenarios involving in-person classes for the University of Michigan’s College of Engineering. We estimate risks of infection,

TB07 CC Room 303B In Person: Business Analytics for Disaster Management General Session Chair: Alfonso Pedraza-Martinez, Indiana University, Indiana University 1 - Inventory Pre-positioning Decision Support for Humanitarian Relief in Nepal Jason Acimovic, Penn State University, University Park, PA, United States Humanitarian organizations typically work independently to pre-position stock in countries in preparation for a disaster. This lack of coordination leads to gaps and overlaps. Working with the Emergency Supply Prepositioning Strategy Group (ESUPS), we utilize inventory data they collected from various NGOs working in Nepal to formulate and solve an optimization problem suggesting where inventory should go. From this, we and ESUPS developed a decision support tool to help guide local NGOs working in Nepal. We will talk about the experience of implementing optimization problem solutions in actual humanitarian organizations, and the status of the project. 2 - Business Analytics for Disaster Management: Research Opportunities And Challenges Alfonso J. Pedraza-Martinez, Indiana University, Kelley School Of Bus., Bloomington, IN, 47405-5308, United States, Lu (Lucy) Yan, Yu Kan We discuss the state of the art of academic and practitioner business-analytics applications for disaster management. In doing so, we identify opportunities for future research in this area. Moreover, we present an empirical application that exemplifies our proposed agenda. TB09 CC Room 303D In Person: Optimization and Surrogate Methods for Black-Box Systems General Session Chair: Hadis Anahideh, University of Illinois at Chicago, Chicago, IL, 60654-4907, United States 1 - An Exploration-Exploitation Approach for Surrogate Optimization Nazanin Nezami, University of Illinois at Chicago, Chicago, IL, United States, Hadis Anahideh The exploration-exploitation trade-off has a critical role in surrogate optimization of expensive black-box functions. Despite the effort of related research in developing strategies to comply with this trade-off, they come short in providing effective guarantees. Proposing a fundamentally different approach to balance this trade-off, we introduce Determinantal Point Processes (DPP) in surrogate optimization. DPP guarantees diversity in a selected subset yet can incorporate the quality of the candidates. This makes DPP a natural choice to balance the exploration-exploitation trade-off. Incorporating the quality component, however, requires careful considerations which is a primary goal of this project. DPP can be successfully utilized for generating representative candidates as well as selecting informative subsets for expensive evaluations. 2 - Bayesian Optimization is Superior to Random Search for Machine Learning Hyperparameter Tuning: Analysis of the Black-box Optimization Challenge 2020 Ryan Turner, Twitter, San Francisco, CA, United States, David Eriksson, David Eriksson, Michael McCourt, Juha Kiili, Eero Laaksonen, Zhen Xu, Isabelle Guyon This paper presents the results and insights from the black-box optimization (BBO) challenge at NeurIPS 2020. The challenge emphasized the importance of evaluating derivative-free optimizers for tuning the hyperparameters of machine learning models. This was the first black-box optimization challenge with a machine learning emphasis. It was based on tuning (validation set) performance of standard machine learning models on real datasets. This competition has widespread impact as black-box optimization is relevant for hyperparameter tuning in almost every machine learning project as well as many applications outside of machine learning. The final leaderboard was determined using the optimization performance on held-out (hidden) objective functions. Baselines were set using several open-source black-box optimization packages as well as random search.

101