When
Thursday, February 27, 2025 - 2:00 p.m.
Enver Yücesan
Professor of Operations Research
INSEAD, France
"Ranking & Selection with Pairwise Comparisons"
ENGR Room 301
Abstract: We consider fixed-budget ranking and selection (R\&S) problems where the performance of alternative designs can only be assessed through pairwise comparisons, a setting encountered in many applications, including player ranking in games, sports tournaments, recommender systems, image-based search, public choice models such as voting schemes or decision rules in committees, and market research. We study two versions of the problem: the first setting is where the pairwise evaluation results in a binary outcome; the performance of an alternative is then quantified by its average probability of “beating” other alternatives. We discuss the challenges associated with applying conventional R&S techniques to this setting and propose heuristic algorithms based on Thompson sampling to overcome those difficulties. The second setting is where the pairwise evaluation results in a real-valued outcome, indicating the score with which one alternative “beat” the other. Assuming Gaussian sampling noise, we extend the knowledge gradient (KG) procedure to adaptively allocate the fixed sampling budget for identifying the best design (or the top-m designs). The proposed algorithms inherit several attractive features of KG such as enabling exact computation and achieving asymptotic mean-variance trade-offs. Finally, we perform numerical experiments to demonstrate that the proposed algorithms deliver competitive and robust finite-budget performance compared with several other state-of-the-art procedures.
(Joint work with Dongyang Li, National University of Singapore, and Chun-Hung Chen, George Mason University)
Bio: Enver Yücesan holds the Abu Dhabi Commercial Bank Chair in International Management in the Technology and Operations Management Area at INSEAD, currently visiting the Department of Supply Chain Management of WP Carey School of Business at Arizona State University. He is an industrial engineer from Purdue University with a PhD in operations research from Cornell University. His research is at the interface of simulation, optimization, and statistics. More specifically, he focuses on complementing the modelling power of computer simulation with efficient analysis methodologies to study the dynamic behavior of complex systems such as supply chains and social networks, which, in turn, enables robust design and effective management of these ecosystems. More recently, he has been focusing on agricultural supply chains to address key challenges such as identification of robust parent seeds, farmer contracting, small holder management, and production and inventory planning under increasing volatility driven by population dynamics and climate change. Over the past three decades, he has also been actively serving the simulation community at large in various editorial and administrative positions; in recognition of his contributions, the INFORMS Simulation Society recently presented Enver with its Distinguished Service Award. He has recently been elected as a Fellow of the Institute for Operations Research and Management Science (INFORMS).