A reinforcement-learning agent learns through trial and error, by interacting with its environment and observing the effect of its actions and the reward that it receives after each action. The reinforcement learning framework is a remarkably powerful one for solving sequential decision tasks and is a major area of research within the field of machine learning. For practical purposes, it is useful if an agent can generalise from tasks it has solved to new but similar tasks it might encounter in the future. Matthijs Snel investigates two classes of strategies for generalization in reinforcement learning.
M. Snel: Generalization Strategies in Reinforcement Learning.
Prof. B.J.A. Kröse.
Prof. F.C.A. Groen
Dr S.A. Whiteson (University of Oxford)