Panel: Memory in Reinforcement Learning
Panel, Conference on Robot Learning – RemembeRL Workshop, Seoul, South Korea
Panel discussion on memory in reinforcement learning.
Talks
Conferences & Workshops
Understanding and Benchmarking Memory in RL
Talk, Conference on Robot Learning – RemembeRL Workshop, Seoul, South Korea
Invited talk on diagnosing and benchmarking memory in reinforcement learning.
Value-Based Reinforcement Learning
Talk, Cambridge ELLIS Unit Summer School on Probabilistic Machine Learning, Cambridge, UK
A 1.5 hour tutorial on value-based reinforcement learning.
Invited Talks
Deep Reinforcement Learning with Memory
Talk, Emotional AI Forum at the University of Macau, Macau SAR, China
Agents trained with deep reinforcement learning can benefit from long-term memory, improving reasoning capabilities just as it does in humans. In this talk, I will give an overview of my research on combining long-term memory with deep reinforcement learning, and some recent and surprising findings suggesting the emergence of fear in our memory-endowed agents.
Reinforcement Learning under Partial Observability
Talk, University of Macau, Macau SAR, China
Reinforcement learning agents have mastered complex games like Go, Atari, or DotA 2, outperforming the best humans on Earth. Yet, these impressive feats are constrained to games or simulations. What prevents us from training superintelligent robots using reinforcement learning? In this talk, we will explore one major roadblock: limited and imperfect sensor data. We will investigate deep memory models as a solution to this challenge. Our journey will cover a number of deep learning architectures, such as recurrent neural networks, graph neural networks, and linear transformers.
An Introduction to Reinforcement Learning
Talk, Toshiba Research, Cambridge, UK
An hour-long talk deriving the basics of value-based and policy-based reinforcement learning.
Guest Lectures
Deep Decision Making
Guest Lecture, University of Macau STGC 8003, Macau SAR, China
Guest lecture on deep decision making.
Reinforcement Learning with Memory
Guest Lecture, University of Cambridge R271, Cambridge, UK
Guest lecture on reinforcement learning with memory.