Hado van Hasselt
Papers
1
Total Citations
184
H-Index
1
About
Hado van Hasselt is a leading researcher in reinforcement learning (RL), best known for pioneering advances in deep RL algorithms and transfer learning. His key contributions include the development of successor features, a framework that enables agents to generalize across tasks by separating dynamics from reward functions, as detailed in his highly cited 2016 work (184 citations). This approach allows for efficient transfer when environments share dynamics but differ in goals, significantly improving sample efficiency. Van Hasselt is also renowned for inventing Double Q-learning, a technique that reduces overestimation bias in value-based RL, which became foundational for algorithms like Double DQN. His work has garnered thousands of citations, reflecting profound impact on both theory and practice—from game-playing agents to robotics. A senior research scientist at DeepMind, van Hasselt’s research continues to shape modern RL, emphasizing robust, scalable methods that bridge the gap between tabular and deep learning paradigms. His contributions are essential reading for anyone seeking to understand how agents can learn faster and more reliably across changing tasks.
Research Focus
Key Achievements
Top Papers
- 1Successor Features for Transfer in Reinforcement Learning184 citations · 2016