Research Talk: Zifeng Zhao (Notre Dame)

Research Talk with Zifeng Zhao (Notre Dame)
Thursday, May 14
11:00am – 12:00pm
Data Science Institute, Room 353
5460 S University Ave

Title: Contextual Dynamic Pricing: Algorithms, Optimality, and Local Differential Privacy Constraints

Abstract: We study contextual dynamic pricing problems where a firm sells products to T sequentially-arriving consumers, behaving according to an unknown demand model. The firm aims to minimize its regret over a clairvoyant that knows the model in advance. The demand follows a generalized linear model (GLM), allowing for stochastic feature vectors in R^d encoding product and consumer information. We first show the optimal regret is of order sqrt{dT}, up to logarithmic factors, improving existing upper bounds by a sqrt{d} factor, achieved by an explore-then-commit (ETC) algorithm. We further study contextual dynamic pricing under local differential privacy(LDP) constraints. We propose a stochastic gradient descent-based ETC algorithm achieving regret upper bounds of order d*sqrt{T}/epsilon, up to logarithmic factors, where epsilon>0 is the privacy parameter. The upper bounds with and without LDP constraints are matched by newly constructed minimax lower bounds, characterizing costs of privacy. Moreover, we extend our study to dynamic pricing under mixed privacy constraints, which naturally bridges private and non-private dynamic pricing. We propose a two-stage ETC algorithm and show that it improves the privacy-utility tradeoff by efficiently leveraging public data. Its optimality is further established via a newly-derived minimax lower bound. To our knowledge, this is the first time such setting is studied in the dynamic pricing literature. Extensive numerical experiments and real data applications are conducted to illustrate the efficiency and practical value of our algorithms.

Bio: Zifeng Zhao is an Associate Professor of Analytics at the Mendoza College of Business, University of Notre Dame. He received B.S. in Financial Risk Management from the Chinese University of Hong Kong, and M.S. in Computer Science and Ph.D. in Statistics from the University of Wisconsin-Madison. He is broadly interested in statistical methodologies and theories on change-point analysis, online learning/bandit, copula, time series and extreme value theory, and their applications in revenue management, risk modeling, and insurance pricing. His research has been funded by the National Science Foundation, Oracle Labs and Kemper Foundation.