TalkRL: The Reinforcement Learning Podcast Martin Riedmiller
Aug 22, 2023
Martin Riedmiller, a research scientist and team lead at DeepMind, discusses using reinforcement learning to control the magnetic field in a fusion reactor. They explore challenges in the TOCOMAC project, reward design, designing actor and critic networks, DQN and NFQ algorithms, the importance of explainability in RL systems, and the horde architecture for collecting experience.
Chapters
Transcript
Episode notes
1 2 3 4 5 6 7 8
Introduction
00:00 • 2min
Previous Control System and the Search for a Better Method
01:58 • 2min
Challenges and Risks of the TOCOMAC Project
03:40 • 8min
Reward Design and Iteration Process
12:00 • 3min
Designing Actor and Critic Networks in Reinforcement Learning
14:57 • 19min
DQN and NFQ: Revolutionizing Reinforcement Learning
33:34 • 18min
The Importance of Explainability in RL Systems
51:57 • 8min
The Horde Architecture and Collecting Experience
01:00:18 • 14min