Prof. Wenpin Tang
Columbia University
Talk: Simulation Optimization with Correlated Replicas and Adaptive Trajectory Evaluations
Abstract: The field of simulation optimization (SO) encompasses various methods developed to optimize complex, expensive-to-sample stochastic systems. In this talk, I will discuss a novel two-stage procedure that leverages LLMs to automate the design of tailored SO algorithms. The first stage constructs an ensemble of digital replicas of the real system. An LLM is employed to implement causal discovery from a textual description of the system, generating a structural ‘skeleton’ that guides the sample efficient learning of the replicas. In the second stage, this replica ensemble is used as an inexpensive testbed to evaluate a set of baseline SO algorithms. An LLM then acts as a meta-optimizer, analyzing the performance trajectories of these algorithms to iteratively revise and compose a final, hybrid optimization schedule. I will also describe an application to blockchain digital twin construction.
Biography: Wenpin Tang is an assistant professor in the Department of Industrial Engineering and Operations Research at Columbia University. Before joining Columbia IEOR, he was a postdoctoral researcher at Department of Industrial Engineering and Operations Research, UC Berkeley and an assistant adjunct professor in the Department of Mathematics at UCLA. He received his PhD in Statistics from UC Berkeley in 2017 and earned his engineering degree from Ecole Polytechnique in 2013. His research interests include probability theory, machine learning, and financial technology. His current projects focus on diffusion generative models and decentralized finance. He has published papers in top journals including the Annals of Probability, Transactions of the American Mathematical Society, and the SIAM Journal on Control and Optimization. He received the Prize for Excellence in Financial Markets from Morgan Stanley in 2017.
