Presented by Jakob Foerster from University of Oxford, this webinar discusses deep reinforcement learning in multi-agent settings and examines how learning occurs in situations where multiple agents are learning at the same time in a decentralized fashion. It is focused on “Learning with Opponent-Learning Awareness” (LOLA), a novel multi-agent reinforcement learning method that allows one agent to consider the learning dynamics of another agent. We show that this not only stabilizes learning in multi-agent settings, but also leads to emergence of cooperation.
By watching this webinar replay, you'll learn:
- how to apply reinforcement learning to multi-agent systems
- the challenges of reciprocity in multi-agent reinforcement learning; and
- why “Learning with Opponent Learning Awareness” (LOLA) can help AI systems learn to navigate areas of conflicting interests