Hado van Hasselt
Also known as H. van Hasselt, Hado Hasselt, Hado P. van Hasselt
Hado van Hasselt is a reinforcement learning researcher whose work spans Q-learning, deep reinforcement learning, and meta-learning. He published a 2010 Utrecht University thesis on temporal-difference learning and joined DeepMind in 2014 after a postdoc with Rich Sutton at the University of Alberta. His 2010 Double Q-learning work reduced action-value overestimation; his 2016 dueling network separated state values from action advantages, and his 2018 PopArt work improved IMPALA’s median score across 57 Atari games. In 2022, he coauthored Bootstrapped Meta-Learning, which received an ICLR Outstanding Paper Award, and he coauthored a 2025 Nature paper on automated discovery of reinforcement-learning algorithms. As of October 2026, he is a research scientist at Google DeepMind.
- Now at
- Google DeepMind
- Works
- 113
- Citations
- 18,101
- h-index
- 30
Publication numbers from OpenAlex, as of 15 October 2026.
Timeline
-
2010
A thesis on temporal-difference learning
He publishes a Utrecht University thesis on temporal-difference learning in 2010.
-
2010
Double Q-learning
He introduces the off-policy Double Q-learning algorithm to reduce overestimation in action values in 2010.
-
2014
Joining DeepMind
After a postdoc with Rich Sutton at the University of Alberta, he moves back to Europe and joins DeepMind in 2014.
-
2016
An ICML best paper
He coauthors Dueling Network Architectures for Deep Reinforcement Learning, which wins the ICML Best Paper Award in 2016; its network separates state values from action advantages.
-
2018
PopArt for Atari agents
In 2018, he and Matteo Hessel show that PopArt improves IMPALA’s median score across 57 Atari games.
-
2022
Bootstrapped meta-learning
He coauthors Bootstrapped Meta-Learning, which receives an ICLR 2022 Outstanding Paper Award.
-
2025
Discovering RL algorithms
He coauthors the 2025 Nature paper Discovering state-of-the-art reinforcement learning algorithms, which reports a machine-discovered RL rule that outperforms manually designed rules.
-
As of October 2026
Research at Google DeepMind
As of October 2026, he is a research scientist at Google DeepMind with research interests in reinforcement learning.
Sources and credits
Photo: We have no published photo of Hado van Hasselt yet. Send us a free-licensed photo.
- Insights in reinforcement ;learning: Formal analysis end empirical evaluation of temporal-difference learning - the University of Groningen research portal
- Centrum Wiskunde & Informatica: Double Q-learning
- Hado van Hasselt Seminar at the Department of Engineering Science, University of Oxford
- Awards | ICML New York City
- Preserving Outputs Precisely while Adaptively Rescaling Targets — Google DeepMind
- Announcing the ICLR 2022 Outstanding Paper Award Recipients — ICLR Blog
- Discovering state-of-the-art reinforcement learning algorithms | Nature
- About me – Hado van Hasselt