Hado van Hasselt
Research scientist at Google DeepMind working on reinforcement learning; first author of Double DQN
- Research scientist, Google DeepMind1
Hado van Hasselt is a research scientist at Google DeepMind whose research interests center on reinforcement learning, and who previously worked with Rich Sutton at the University of Alberta. Van Hasselt is first author of "Deep Reinforcement Learning with Double Q-learning" (2015), which shows that the DQN algorithm suffers from overestimated action values and proposes an adaptation to reduce them. Van Hasselt's site also documents teaching the reinforcement learning part of UCL's Advanced Topics in Machine Learning course in 2016, with Joseph Modayil.123
Work
Double DQN (external site: arxiv.org)Published research
First author, with Arthur Guez and David Silver, of the 2015 paper on overestimation in DQN and its double Q-learning adaptation.3
This profile covers public professional work only, from the sources below. It is not an endorsement or a ranking. To correct or remove it, use the correction form.