INFORMS 2021 Program Book

INFORMS Anaheim 2021

TB10

3 - Smart-Replication for Black-Box Optimization under Uncertainty Hadis Anahideh, University of Illinois at Chicago, Chicago, IL, 60654-4907, United States, Jay Michael Rosenberger, Victoria C. P. Chen Optimizing high-dimensional expensive black-box systems under uncertainty is an extremely challenging problem. As a resolution, we develop a novel replication approach called Smart-Replication to overcome the uncertainties associated with the black-box output. The Smart-Replication approach identifies promising input points to replicate and avoids unnecessary evaluations of other data points. It is agnostic to the choice of a surrogate and can adapt itself to an unknown noise level. We demonstrate the effectiveness of the Smart-Replication approach using interpolating and non-interpolating surrogates on different complex global optimization test functions. TB10 CC Room 304B In Person: Data Analytics in Agriculture General Session Chair: Fatemeh Amini, Iowa State University, Ames, IA, 50014, United States 1 - Sparse Testing (ST) in Plant Breeding Reka Howard, University of Nebraska-Lincoln, Lincoln, NE, 68583, United States Sparse testing (ST) in plant breeding is the situation where not all genotypes of interest are grown in each environment. Using genomic prediction and genotype x environment interaction (GE), the non-observed genotype-in-environment combinations can be predicted. The accuracy of predicting the unobserved data depends on (1) how many genotypes overlap between environments, (2) in how many environments each genotype is grown, and (3) which prediction method is used. Here, we studied the predictive ability obtained when using a fixed number of plots and different ST designs. The empirical study was based on maize hybrid data collected in three environments. Three different prediction models were implemented, two main effects models, and a model including the GE term. The GE model had higher prediction accuracy than the other models for the different allocation scenarios. 2 - Agricultural Genome to Phenome Initiative: Shared Data Science Across Crop and Livestock Communities Jennifer Clarke, University of Nebraska-Lincoln, Lincoln, NE, United States, Jack Dekkers, Carolyn Lawrence-Dill, Eric Lyons, Brenda Murdoch, Patrick Schnable, Christopher Tuggle To achieve sustainable genetic improvements of agricultural species, the expertise of a broad community of agricultural researchers must be engaged from both crop and livestock communities. This includes integrative disciplines such as statistics and the data and engineering sciences. The objective of the Agricultural Genome to Phenome Initiative (AG2PI) is to assemble and prepare a transdisciplinary community to conduct AG2P research. To accomplish this, AG2PI seeks to engage a broad and diverse researcher community through Field Days, Conferences, Training workshops, and Seed grants. In this presentation we will provide an overview of AG2PI and highlight examples of AG2P data science research. We will provide information about AG2PI activities including seed grants to support projects that will develop and enable FAIR data science in agricultural contexts. 3 - Blockchain Technology and the Sustainable Supply Chain: Theoretically Exploring Adoption Barriers Sara Saberi, Worcester Polytechnic Institute (WPI), Worcester, MA, 1609, United States, Mahtab Kouhizadeh, Joseph Sarkis In this study, the technology-organization-environment framework andforce field theories are utilized to investigate blockchain adoption barriers. Using various literature streams on technology, organizational practices, and sustainability, a comprehensive overview of barriers for adopting blockchain technology to manage sustainable supply chains is provided. The barriers are explored using technology, organizational, and environmental supply chain and external framework followed by inputs from academics and industry experts and then analyzed using the Decision-Making Trial and Evaluation Laboratory (DEMATEL) tool. The results show that supply chain and technological barriers are the most critical barriers among both academics and industry experts. 4 - Nonlinear Multi-objective Optimization Selection Strategy In Multi-trait Genomic Selection Fatemeh Amini, Iowa State University, Ames, IA, 50014-7776, United States, Guiping Hu, Lizhi Wang Genomic Selection mostly aims at using genotypic data to identify elite breeding parents and the mating strategy to enhance only one trait at a time. Although some of the research address multi-trait genomic selection which improves multiple traits simultaneously, they mostly use linear index selection in identifying elite breeding parents. In this paper, we proposed a nonlinear selection approach that adopts a nonlinear multi-objective optimization function which can be adopted with any desired mating strategy. A simulation platform is designed to compare the optimal Pareto Frontier of performance of different selection strategies in the final generation of SAM maize dataset. The results demonstrate

that the nonlinear multi-objective approach outperforms linear index selection, considering both continuous and discrete traits.

TB11 CC Room 304C In Person: Economics and Computation IV Award Session Chair: Ali Aouad, London Business School, London, NW6 4TG, United Kingdom 1 - Fair Dynamic Rationing Scott Rodilitz, Yale, New Haven, CT, United States, Vahideh Manshadi, Rad Niazadeh We study the allocative challenges that governmental and nonprofit organizations face when rationing of a social good among agents whose needs (demands) realize sequentially and are possibly correlated. To better achieve equity and efficiency in such contexts, social planners intend to maximize the minimum fill rate across agents. We show that a simple adaptive policy of projected proportional allocation achieves the best possible expected minimum fill rate (ex- post fairness) and minimum expected fill rate (ex-ante fairness). Our policy is transparent and easy to implement, and we demonstrate its effectiveness with a numerical study motivated by the rationing of COVID-19 medical supplies. 2 - The Limits to Learning a Diffusion Model Andrew Zheng, Massachusetts Institute of Technology, Cambridge, MA, United States, Vivek Farias, Jackie W. Baek, Tianyi Peng, Joshua T. Wilde, Deeksha Sinha, Retsef Levi, Andreea Georgescu We provide the first sample complexity lower bounds for the estimation of simple diffusion models, including the Bass model (for product adoption) and the SIR model (for epidemics). For Bass models with low innovation rates, our results imply that one cannot predict the eventual number of adopting customers until one is at least two-thirds of the way to the time at which the rate of new adopters is at its peak. For the SIR model, one cannot predict the eventual number of infections until one is approximately two-thirds of the way to the time at which the infection rate has peaked. These limits are borne out in both product adoption data (Amazon), as well as epidemic data (COVID-19). 3 - Incomplete Information VCG Contracts for Common Agency Elisheva S. Shamash, PhD student, Technion, Haifa, Israel, Tal Alon, Ron Lavi, Inbal Talgam-Cohen We study contract design for welfare maximization in the ``common agency’’ model [Bernheim and Whinston,1986], coordinating multiple principals with incomplete information of agent’s action. Extending complete-information VCG contracts [Lavi and Shamash, 2019] to incomplete information,we characterize ``incomplete information VCG contracts (IIVCG)’’, and show uniquness guaranteeing truthfulness and welfare maximization. We reveal a tradeoff between individual rationality and limited liability, which insure participation. We design a polynomial-time algorithm determining whether a setting has an IIVCG contract with both properties, and if possible, returning such a contract. 4 - Dynamic Pricing and Learning under the Bass Model Steven Yin, Columbia University, New York, NY, United States, Shipra Agrawal, Assaf Zeevi We consider a novel formulation of the dynamic pricing and demand learning problem, where the evolution of demand in response to posted prices is governed by a stochastic variant of the popular Bass model with parameters $(\alpha, \beta)$ representing the so-called “innovation” and “imitation” effects. In this model the posted price not only affects the demand in the current round but also the future evolution of demand. Our main contribution is the development of an algorithm that satisfies a high probability regret guarantee of order $\tilde O(m^{2/3})$; where the market size $m$ is known a priori. Moreover, we show that no algorithm can incur smaller order of loss by deriving a matching lower bound. 5 - Online Assortment Optimization for Two-sided Matching Platforms Ali Aouad, London Business School, London, United Kingdom, Daniela Saban Motivated by online labor platforms, we study a two-sided online assortment optimization problem. Each customer arrives and requests to match with a supplier out of the displayed assortment; subsequently, suppliers make choice decisions over the set of requests they received. We show that myopic algorithms attain the best-possible competitive ratio for this problem under general rank- based choice models. We devise “preference-aware” balancing algorithms that achieve higher competitive ratios under logit-based choice models. Interestingly, the performance of online algorithms is tightly connected to the structure of the supplier-side choice model.

102