• Sorted by Date • Classified by Publication Type • Classified by Research Category •
Marc Ponsen, Pieter Spronck, and Karl Tuyls. Towards Relational Hierarchical Reinforcement Learning in Computer Games. In Proceedings of the 18th Benelux Conference on Artificial Intelligence (BNAIC 2006), October 5-6, Namur, Belgium, 2006.
Computer games are challenging test beds for machine learning research.Without applying abstraction and generalization techniques, many traditionalmachine learning techniques, such as reinforcement learning, will failto learn efficiently. In this paper we examine extensions of reinforcementlearning that scale to the complexity of computer games. In particular welook at hierarchical reinforcement learning applied to a learning task in areal time-strategy computer game. Moreover, we provide a first step towardsrelational reinforcement learning in computer games by introducing arelational representation of the state features and actions. We found that hierarchicalreinforcement learning significantly outperforms flat reinforcementlearning for our task. We also show that a relational state representationallows the adaptive agent to learn a generalized policy, i.e., it is capable oftransferring knowledge to unseen task instances.
@INPROCEEDINGS{Ponsbnaic06,
author = {Marc Ponsen and Pieter Spronck and Karl Tuyls},
title = {Towards Relational Hierarchical Reinforcement Learning in Computer
Games},
booktitle = {Proceedings of the 18th Benelux Conference on Artificial Intelligence
(BNAIC 2006)},
year = {2006},
abstract = {Computer games are challenging test beds for machine learning research.
Without applying abstraction and generalization techniques, many traditional
machine learning techniques, such as reinforcement learning, will fail
to learn efficiently. In this paper we examine extensions of reinforcement
learning that scale to the complexity of computer games. In particular we
look at hierarchical reinforcement learning applied to a learning task in a
real time-strategy computer game. Moreover, we provide a first step towards
relational reinforcement learning in computer games by introducing a
relational representation of the state features and actions. We found that hierarchical
reinforcement learning significantly outperforms flat reinforcement
learning for our task. We also show that a relational state representation
allows the adaptive agent to learn a generalized policy, i.e., it is capable of
transferring knowledge to unseen task instances.},
address = {October 5-6, Namur, Belgium},
bib2html_pubtype = {Refereed Conference},
bib2html_rescat = {Multi-Agent Learning},
owner = {K.Tuyls},
timestamp = {2006.11.23},
}
Generated by bib2html.pl (written by Patrick Riley ) on Mon Jan 18, 2010 09:51:52