Episode · Dec 26, 2021
BI 123 Irina Rish: Continual Learning
Brain Inspired · 1 h 19 min
We use cookies for analytics.
Doina Precup discusses how hierarchical reinforcement learning lets agents learn abstractions across time. She examines how much a reward signal can express, including results on Markov rewards, and connects temporal abstraction with the problem of agents that continue learning as their environment changes.