Abstract
Many reinforcement learning domains are highly relational.Whiletraditional temporal-difference methods can be applied to these domains, they are limited in their capacity to exploit the relational nature of the domain. Our algorithm, AMBIL, constructs relational world models in the form of relational Markov decision processes (MDPs). AMBIL works backwards from collections of high-reward states, utilizing inductive logic programming to learn their preimage, logical definitions of the region of state space that leads to the high-reward states via some action. These learned preimages are chained together to form an MDP that abstractly represents the domain. AMBIL estimates the reward and transition probabilities of this MDP from past experience. Since our MDPs are small, AMBIL uses value-iteration to quickly estimate the Q-values of each action in the induced states and determine a policy. AMBIL is able to employ complex background knowledge and supports relational representations. Empirical evaluation on both synthetic domains and a sub-task of the RoboCup soccer domain shows significant performance gains compared to standard Q-learning.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
Similar content being viewed by others
References
Bellman, R.E.: Dynamic Programming. Princeton University Press, Princeton, New Jersey (1957)
Blockeel, H., De Raedt, L.: Top-down induction of first-order locical decision trees. Artificial Intelligence (June 1998)
Dietterich, T., Flann, N.: Explanation-based learning and reinforcement learning: A unified view. In: Proceedings of the International Conference on Machine Learning (1995)
Driessens, K., Ramon, J., Blockeel, H.: Speeding up relational reinforcement learning through the use of an incremental first order decision tree algorithm. In: Proceedings of the European Conference on Machine Learning (2001)
Džeroski, S., De Raedt, L., Blockeel, H.: Relational reinforcement learning. In: Proceedings of the International Conference on Machine Learning (1998)
Lecoeuche, R.: Learning optimal dialog management rules by using reinforcement learning and inductive logic programming. In: Proceedings of the North American Chapter of the Association of Computational Linquistic (June 2001)
Kersting, K., Van Otterlo, M., De Raedt, L.: Bellman goes relational. In: Proceedings of the International Conference on Machine Learning (2004)
Maclin, R., Shavlik, J., Torrey, L., Walker, T., Wild, E.: Giving advice about preferred actions to reinforcement learners via knowledge-based kernel regression. In: Proceedings of the Twentieth Conference on Artificial Intelligence (2005)
Morales, E.F.: Scaling up reinforcement learning with a relational representation. In: Proceedings of the Workshop on Adaptability in Multi-Agent Systems at AORC 2003 (2003)
Quinlan, J.R.: C4.5: Programs for Machine Learning. Morgan Kaufmann, San Francisco (1993)
Van Otterlo, M.: Efficient reinforcement learning using relational aggregation. In: Proceedings of the Sixth European Workshop on Reinforcement Learning (2003)
Srinivasan, A.: The Aleph Manual (2001)
Stone, P., Sutton, R.: Scaling reinforcement learning toward RoboCup soccer. In: Proceedings of the International Conference on Machine Learning (2001)
Sutton, R.: Integrated modeling and control based on reinforcement learning and dynamic programming. In: Advances in Neural Information Processing Systems, vol. 3 (1991)
Sutton, R., Barto, A.: Reinforcement Learning: An Introduction. MIT Press, Cambridge (1998)
Watkins, C.J.C.H.: Learning from Delayed Rewards. Ph.D. thesis, Cambridge University (1989)
Zettlemoyer, L.S., Pasula, H.M., Kaelbling, L.P.: Learning Planning Rules in Noisy Stochastic Worlds. In: Proceedings of the Twentieth Conference on Artificial Intelligence (2005)
Author information
Authors and Affiliations
Editor information
Rights and permissions
Copyright information
© 2008 Springer-Verlag Berlin Heidelberg
About this paper
Cite this paper
Walker, T., Torrey, L., Shavlik, J., Maclin, R. (2008). Building Relational World Models for Reinforcement Learning. In: Blockeel, H., Ramon, J., Shavlik, J., Tadepalli, P. (eds) Inductive Logic Programming. ILP 2007. Lecture Notes in Computer Science(), vol 4894. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-78469-2_27
Download citation
DOI: https://doi.org/10.1007/978-3-540-78469-2_27
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-78468-5
Online ISBN: 978-3-540-78469-2
eBook Packages: Computer ScienceComputer Science (R0)