Wrappers for Automatic Parameter Tuning in Multi-Agent
Optimization by Genetic Programming
W. Hsu, and S. Gustafson. IJCAI-2001 Workshop on Wrappers for Performance
Enhancement in Knowledge Discovery in Databases (KDD), Seattle, Washington, USA, (4 August 2001)
Abstract
We present an adaptation of the standard genetic
program (GP) to hierarchically decomposable,
multi-agent learning problems. To break down a problem
that requires cooperation of multiple agents, we use
the team objective function to derive a simpler,
intermediate objective function for pairs of
cooperating agents. We apply GP to optimize first for
the intermediate, then for the team objective function,
using the final population from the earlier GP as the
initial seed population for the next. This layered
learning approach facilitates the discovery of
primitive behaviors that can be reused and adapted
towards complex objectives based on a shared team goal.
We use this method to evolve agents to play a
subproblem of robotic soccer (keep-away soccer).
Finally, we show how layered learning GP evolves better
agents than standard GP, including GP with
automatically defined functions, and how the problem
decomposition results in a significant learning-speed
increase.
IJCAI-2001 Workshop on Wrappers for Performance
Enhancement in Knowledge Discovery in Databases (KDD)
year
2001
month
4 August
notes
http://www.kddresearch.org/KDD/Workshops/IJCAI-2001/
Paper from author 19 Jul 2001. Also available as
GECCO'2001 late breaking paper Coaching, seeding, LLGP,
keep-away soccer (minimize number of turnovers), MAS,
RoboCup, passing agents and keep-away soccer agents,
ADF. Simple GP and ADFGP trained with one shot fitness
function, ie not layered. Popsize 2000 "yeilded good
results". Luke's ECJ. SoccerServer, TeamBots.
See also gustafson:mastersthesis
%0 Conference Paper
%1 hsu:2001:waptmaoGP
%A Hsu, William H.
%A Gustafson, Steven M.
%B IJCAI-2001 Workshop on Wrappers for Performance
Enhancement in Knowledge Discovery in Databases (KDD)
%C Seattle, Washington, USA
%D 2001
%K algorithms, genetic programming, robotic soccer
%T Wrappers for Automatic Parameter Tuning in Multi-Agent
Optimization by Genetic Programming
%X We present an adaptation of the standard genetic
program (GP) to hierarchically decomposable,
multi-agent learning problems. To break down a problem
that requires cooperation of multiple agents, we use
the team objective function to derive a simpler,
intermediate objective function for pairs of
cooperating agents. We apply GP to optimize first for
the intermediate, then for the team objective function,
using the final population from the earlier GP as the
initial seed population for the next. This layered
learning approach facilitates the discovery of
primitive behaviors that can be reused and adapted
towards complex objectives based on a shared team goal.
We use this method to evolve agents to play a
subproblem of robotic soccer (keep-away soccer).
Finally, we show how layered learning GP evolves better
agents than standard GP, including GP with
automatically defined functions, and how the problem
decomposition results in a significant learning-speed
increase.
@inproceedings{hsu:2001:waptmaoGP,
abstract = {We present an adaptation of the standard genetic
program (GP) to hierarchically decomposable,
multi-agent learning problems. To break down a problem
that requires cooperation of multiple agents, we use
the team objective function to derive a simpler,
intermediate objective function for pairs of
cooperating agents. We apply GP to optimize first for
the intermediate, then for the team objective function,
using the final population from the earlier GP as the
initial seed population for the next. This layered
learning approach facilitates the discovery of
primitive behaviors that can be reused and adapted
towards complex objectives based on a shared team goal.
We use this method to evolve agents to play a
subproblem of robotic soccer (keep-away soccer).
Finally, we show how layered learning GP evolves better
agents than standard GP, including GP with
automatically defined functions, and how the problem
decomposition results in a significant learning-speed
increase.},
added-at = {2008-06-19T17:35:00.000+0200},
address = {Seattle, Washington, USA},
author = {Hsu, William H. and Gustafson, Steven M.},
biburl = {https://www.bibsonomy.org/bibtex/22eaf37b4bb49d5ba4c6a85cdc04f9841/brazovayeye},
booktitle = {IJCAI-2001 Workshop on Wrappers for Performance
Enhancement in Knowledge Discovery in Databases (KDD)},
interhash = {e1de601b5c599797a87a0bdfbe7faf80},
intrahash = {2eaf37b4bb49d5ba4c6a85cdc04f9841},
keywords = {algorithms, genetic programming, robotic soccer},
month = {4 August},
notes = {http://www.kddresearch.org/KDD/Workshops/IJCAI-2001/
Paper from author 19 Jul 2001. Also available as
GECCO'2001 late breaking paper Coaching, seeding, LLGP,
keep-away soccer (minimize number of turnovers), MAS,
RoboCup, passing agents and keep-away soccer agents,
ADF. Simple GP and ADFGP trained with one shot fitness
function, ie not layered. Popsize 2000 {"}yeilded good
results{"}. Luke's ECJ. SoccerServer, TeamBots.
See also \cite{gustafson:mastersthesis}},
timestamp = {2008-06-19T17:41:52.000+0200},
title = {Wrappers for Automatic Parameter Tuning in Multi-Agent
Optimization by Genetic Programming},
year = 2001
}