Approximate Exploration through State Abstraction

Adrien Ali Taïga, Aaron Courville, Marc G. Bellemare

Although exploration in reinforcement learning is well understood from a theoretical point of view, provably correct methods remain impractical. In this paper we study the interpla…