Research

My research falls within the area of Artificial Intelligence known as Reinforcement Learning (RL). An artificial Reinforcement Learning agent learns to optimize an engineered "reward" signal by interacting with its environment: an agent which is rewarded for kicking a soccer ball should learn, over time, to be a good soccer ball kicker. This field of research has strong ties with research in neuroscience, which suggests that a similar mechanism may form the basis for reward related dopamine systems in the brain.

My particular research focusses on methods of abstraction in Reinforcement Learning. I look at algorithms that may help an agent decide what to pay attention to and what to ignore in its environment, in particular, how an agent can build multiple abstract views of its environment in order to complete multiple tasks.