Hi, I am William. I work on strategy and skill learning for multi-robot systems.
Training one robot to do a hard task is difficult enough. In a team, each robot also has to work out who does what, predict what the others will do, and keep learning while they are learning too. Sometimes the other agent is not a teammate. It might be a moving obstacle or an opponent, and the robot has to work out what it intends and react to it. I work on these problems using reinforcement learning and game theory.
I am also interested in representation learning, that is, what an agent learns to pay attention to and what it ignores. This affects what a team can coordinate on, and whether a skill learned on one task still works on another.
Learning outside the lab
I skateboard and I learn languages, and I keep running into the same questions there. Why do some hours of practice count for much more than others? How should practice be ordered, and what kind of feedback helps? I think these are the same problems a learning agent faces, just in different words.
Mathematics
I read topology, graph theory, category theory, group theory and geometry for my own interest. I like how the same structure, such as a symmetry, turns up in things that look unrelated.
Publications
My papers are on the publications page.
Writing
This site is also my personal blog, where I write about my interests (which are many) and scratch my childhood itch of being a writer. The posts are mostly about research and mathematics, but not always.