An Introduction to Reinforcement Learning and Optimal Control Theory

Organized by CCM – Chair of Computational Mathematics

Speaker: Carlos Esteve Yague, Postdoctoral Researcher at CCM

From September 8th. to October 1st, 2020

Sessions: 4, one session/week

Abstract. This mini-course aims to be an introduction to Reinforcement Learning for people with a background in control theory. We will discuss the differences and similarities between the two settings, relying on Markov decision processes (MDP) and dynamical systems (DS) respectively. We will present and analyze the most elementary Reinforcement Learning techniques, based on the dynamic programming principle. By means of the HJB equation, we will also discuss the possibility of implementing RL methods in continuous settings. Finally, we will consider inverse problems arising in this context, where the goal is to identify the underlying dynamics of the system and/or a cost functional compatible with a given optimal policy.

Sessions planning:

Date & Room Session | Subject
08.09.2020 10:00-11:30
Google Meet & Logistar Room at DeustoTech
Introduction and Dynamic Programming Methods
17.09.2020 12:00-13:00
Google Meet & Logistar Room at DeustoTech
From discrete to continuous models, Hamilton-Jacobi-Bellman equations
24.09.2020 12:00-13:00
Google Meet & Logistar Room at DeustoTech
Q-Learning
01.10.2020 12:00-13:00
Google Meet & Logistar Room at DeustoTech
TBA