Online learning in two-sided markets
Designing mechanisms that facilitate the exchange of items between two or more strategic agents while maximizing social welfare is a fundamental topic in economics. In particular, bilateral trade -- where only one buyer and one seller are considered-- has been studied in recent years through the lens of regret minimization. The design of two-sided market mechanisms through online learning presents some interesting theoretical challenges primarily due to the nature of the feedback, which can be in principle harder than the ones usually available in other settings. ; In this research, I aim to address open problems in online learning for two-sided markets, explore how relaxing certain requirements -- such as the restriction to only two agents or the budget balance constraint -- affects the achievable regret, and design novel algorithms to better tackle each setting. Additionally, I will examine different assumptions about agents' nature, studying the problem in both stochastic and adversarial environments.
Back to Current Students