Using relative novelty to identify useful temporal abstractions in reinforcement learning.

zgr Simsek, Andrew G. Barto

Browse the full ICML paper archive.