2016 Poster Sessions : Learning and Earning: Dynamic Pricing in an Unknown Environment with Memory

Student Name : Abbas Kazerouni
Advisor : Benjamin Van Roy
Research Areas: Information Systems
We consider a monopolist seller who is trying to maximize his profit by dynamically changing the prices of his items. As a principle in economics, the current literature assumes that the demand for an item at any time a decreasing function of the price at that time. This assumption might be far from reality as it does not capture the effect of shopping sales on the demand for an item. To capture such effects, we develop an extended economical model where the demand for an item depends on both its current price and its previous prices. Since the underlying dependence of the demand on the price of an item is not exactly known to the seller, it needs to be learned by exploring different prices during the interaction with the customers. At the same time, the seller wants to exploit the available knowledge to set the prices to maximize his revenue. To address this exploration-exploitation dilemma, we propose a dynamic pricing strategy which achieves a regret bound of the same order as the lower bound. In the special case of linear demand function, our algorithm achieves the same regret bound as it would be achieved in a memoryless environment.

I am a 3rd year PhD student in EE department at Stanford University being jointly supervised by Lawrence M. Wein and Benjamin Van Roy. My research interest lies at the intersection of Reinforcement Learning and Operation Management with applications in Healthcare and Economics. Specifically, I uses Statistical Learning techniques to design algorithms that can learn to make decisions under uncertainty. Before joining Stanford, I received a BSc in Electrical Engineering and a BSc in Mathematics from Sharif University of Technology in 2013.