Deep Learning and Reinforcement Learning have emerged in recent years as two of the core areas of research in Artificial Intelligence. In this talk we will present the fundamental ideas underlying reinforcement learning algorithms, the specific challenges that emerge when combining Reinforcement Learning and Deep Learning, highlighting open problems and promising directions of research.

Matteo Hessel is a Research Engineer at DeepMind. His research focuses on building general artificial agents, capable of learning to perform a variety of complex tasks. He believes that the combination of Deep Learning techniques with Reinforcement Learning will be a crucial component in order to achieve this. Past work includes specialized deep learning architectures for Reinforcement Learning (Dueling Networks, PopArt and Predictron), and the combination of multiple algorithmic components in a single integrated agent (Rainbow). He loves teaching.