Writing
Research, explained from the inside.
Short essays about the questions, design choices, and intuitions behind my work. These articles complement the formal papers; they do not replace them.
Exploration Needs Disagreement: Noise, Priors, and Bootstrapped DQN
Why ensemble diversity is not a cosmetic property but a mechanism for deep exploration.
Deep reinforcement learningExplorationDQN
What Should a Representation Remember When the World Is Partially Observable?
Partial observability turns representation learning from image compression into an inference problem.
Reinforcement learningPartial observabilityState representation
The Unbalanced Atlas: Why Representation Spaces Need More Than One Local View
An intuitive account of our ICLR 2024 work on manifold-aware self-supervised state representation learning.
Self-supervised learningManifoldsICLR 2024