A novel multi-step reinforcement learning method for solving reward hacking
Sign Up to like & getrecommendations! Published in 2019 at "Applied Intelligence"
DOI: 10.1007/s10489-019-01417-4
Abstract: Reinforcement learning with appropriately designed reward signal could be used to solve many sequential learning problems. However, in practice, the reinforcement learning algorithms could be broken in unexpected, counterintuitive ways. One of the failure modes… read more here.
Keywords: reward hacking; multi step; method; reinforcement learning ... See more keywords