On September 11th, 2020 Lorenzo Bisi will take a talk on “Robust Reinforcement Learning” as a part of his Summer School ACACES 2020.

Robustness has revealed to be a critical aspect for the applicability of Supervised Learning techniques to real-world problems. Do the same issues affect Reinforcement Learning (RL) too? In this talk, different concepts of robustness for RL will be formalized, and we will understand when it is important for an RL application to be robust. The Robust MDP framework will be described, analysing the more effective solution approaches and theoretical results. Finally, the main research lines will be illustrated, and we will provide a connection with the Risk-Averse RL literature.

The event will be hold online via Microsoft Teams.