Abstract

In this talk I will present my recent works on resource constrained revenue management with demand learning and large action spaces. We study a class of well-known RM problems such as dynamic pricing and assortment optimization subject to non-replenishable inventory constraints, where demand or choice model information is unknown a priori and needs to be estimated, and the action spaces (price vectors, assortments) are large. We present a general primal-dual optimization algorithm with upper confidence bounds to achieve optimal asymptotic regret. We also extend this result to nonparametric demand modeling in network revenue management problems via a robust ellipsoid method.

Paper links:

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3841273

https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3948140

Speaker

Yining Wang is an associate professor of operations management at Naveen Jindal School of Management, University of Texas at Dallas. He graduated with a PhD in Machine Learning from Carnegie Mellon University. His research primarily focuses on machine learning and online learning methodology with applications in operations and revenue management. He is also interested in ethics questions arising from the use of machine learning and AI in personalized revenue management systems, such as data privacy protection and decision fairness issues.