Reinforcement Learning: Frontiers

This class covers subjects of contemporary research contributing to the design of reinforcement learning agents that can operate effectively across a broad range of environments. Topics include exploration, generalization, credit assignment, and state and temporal abstraction. An important component of the class is a research project aimed at understanding a focused issue in reinforcement learning. Can be repeated for credit. Prerequisites: 226, CS 234, or EE 277, and experience with mathematical proofs.