GT « Analyse, Algorithmique, Apprentissage »

Sasha Rakhlin (MIT)

Title:

On the Foundations of Interactive Decision Making

Abstract:

We present a general framework for interactive decision making that subsumes multi-armed bandits, contextual bandits, structured bandits, and reinforcement learning. In these settings, the statistician interacts with the environment to collect data and necessarily faces an exploration-exploitation dilemma. We focus on the statistical aspect of learning in this interactive setting, aiming to develop a tight characterization of sample complexity in terms of properties of the class of models.