Theoretical and Empirical Analysis of Reward Shaping in Reinforcement Learning

TY - CONF

T1 - Theoretical and Empirical Analysis of Reward Shaping in Reinforcement Learning

AU - Grzes, Marek

AU - Kudenko, Daniel

PY - 2009

Y1 - 2009

N2 - Reinforcement learning suffers scalability problems due to the state space explosion and the temporal credit assignment problem. Knowledge-based approaches have received a significant attention in the area. Reward shaping is a particular approach to incorporate domain knowledge into reinforcement learning. Theoretical and empirical analysis of this paper reveals important properties of this principle, especially the influence of the reward type, MDP discount factor, and the way of evaluating the potential function on the performance.

AB - Reinforcement learning suffers scalability problems due to the state space explosion and the temporal credit assignment problem. Knowledge-based approaches have received a significant attention in the area. Reward shaping is a particular approach to incorporate domain knowledge into reinforcement learning. Theoretical and empirical analysis of this paper reveals important properties of this principle, especially the influence of the reward type, MDP discount factor, and the way of evaluating the potential function on the performance.

UR - http://www.scopus.com/inward/record.url?scp=77950814349&partnerID=8YFLogxK

M3 - Paper

SP - 337

EP - 344

ER -