Skip to contentSkip to filters

We use cookies for analytics.

Rethinking the Foundations for Continual Reinforcement Learning

Michael Bowling

University of Alberta

Recording

48 min

Abstract

Michael Bowling examines whether the usual foundations of reinforcement learning suit agents that need to adapt indefinitely. He discusses how concepts such as optimal policies and Markov decision processes can limit the questions researchers ask about changing environments and proposes alternative foundations as a starting point for new algorithms.

Topics