Vidal's libraryTitle: | Multi-agent learning in extensive games with complete information |
Author: | Pu Huang and Katia Sycara |
Book Tittle: | Proceedings of the second international joint conference on Autonomous agents and multiagent systems |
Pages: | 701--708 |
Publisher: | ACM Press |
Year: | 2003 |
DOI: | 10.1145/860575.860688 |
Abstract: | Learning in a multi-agent system is di cult because the learning environment jointly created by all learning agents is time-variant. This paper studies the model of multi-agent learning in complete-information extensive games (CEGs). We provide two provably convergent algorithms for this model. Both algorithms utilize the special structure of CEGs and guarantee both individual and collective convergence. Our work contributes to the multi-agent learning literature in several aspects: 1. We identify a model of multi-agent learn- ing, namely, learning in CEGs, and provide two provably convergent algorithms for this model. 2. We explicitly address the environment-shifting problem and show that how patient agents can collectively learn to play equilib- rium strategies. 3. Many game-theoretical work on learning uses a technique called fictitious play, which requires agents to build beliefs about their opponents. For our model of learning in CEGs, we show it is true that agents can collec- tively converge to the sub-game perfect equilibrium (SPE) by repeatedly reinforcing their previous success/failure ex- perience; no belief building is necessary. |
Cited by 2 - Google Scholar
@inproceedings{huang03a,
author = {Pu Huang and Katia Sycara},
title = {Multi-agent learning in extensive games with
complete information},
booktitle = {Proceedings of the second international joint
conference on Autonomous agents and multiagent
systems},
year = 2003,
pages = {701--708},
location = {Melbourne, Australia},
doi = {10.1145/860575.860688},
publisher = {ACM Press},
abstract = {Learning in a multi-agent system is di cult because
the learning environment jointly created by all
learning agents is time-variant. This paper studies
the model of multi-agent learning in
complete-information extensive games (CEGs). We
provide two provably convergent algorithms for this
model. Both algorithms utilize the special structure
of CEGs and guarantee both individual and collective
convergence. Our work contributes to the multi-agent
learning literature in several aspects: 1. We
identify a model of multi-agent learn- ing, namely,
learning in CEGs, and provide two provably
convergent algorithms for this model. 2. We
explicitly address the environment-shifting problem
and show that how patient agents can collectively
learn to play equilib- rium strategies. 3. Many
game-theoretical work on learning uses a technique
called fictitious play, which requires agents to
build beliefs about their opponents. For our model
of learning in CEGs, we show it is true that agents
can collec- tively converge to the sub-game perfect
equilibrium (SPE) by repeatedly reinforcing their
previous success/failure ex- perience; no belief
building is necessary.},
keywords = {multiagent game-theory learning},
url = {http://jmvidal.cse.sc.edu/library/huang03a.pdf},
googleid = {o1RUIWpozS0J:scholar.google.com/},
comment = {masrg},
cluster = {3300408906967438499},
}
Last modified: Wed Mar 9 10:15:44 EST 2011