My current research interests involve the recursive decomposition of complex, long-horizon tasks into simpler subproblems. I view hierarchical learning as both a structural prior that improves optimization in reinforcement learning and a framework for continual learning of increasingly sophisticated skills.

Previously, I also spent some time studying computational neuroscience, investigating how the brain allocates control between model-free and model-based decision-making strategies and developing neural decoding methods for brain-computer interfaces.

If you're also interested in these topics, feel free to reach out to john.ly.zhou at gmail.com!

Blog

Research

Previous Research