DocumentCode :
2820274
Title :
Evolution of strategies in a spatial IPD game with a number of different representation schemes
Author :
Ishibuchi, Hisao ; Hoshino, Koichiro ; Nojima, Yusuke
Author_Institution :
Dept. of Comput. Sci. & Intell. Syst., Osaka Prefecture Univ., Sakai, Japan
fYear :
2012
fDate :
10-15 June 2012
Firstpage :
1
Lastpage :
8
Abstract :
We examine the evolution of strategies for a spatial IPD (Iterated Prisoner´s Dilemma) game, which are encoded using different representation schemes. Each agent at a cell in a two-dimensional grid-world has its own representation scheme for encoding its strategy. In general, strategies with different representation schemes cannot be recombined. Thus a population of agents can be viewed as a mixture of different species (i.e., an ecology with different species). When the size of a neighborhood structure is small and/or the number of representation schemes is large, it is likely that some agents have no neighbors with the same representation scheme. We discuss the handling of those agents because they cannot generate their new strategies through recombination. In computational experiments, we use four types of strings (i.e., four representation schemes). Agents in our spatial IPD game are randomly divided into four groups of the same size. One string type is assigned to each group. Recombination is performed between strings of neighboring agents with the same string type. With respect to the IPD game, we compare two settings with each other. In one setting, the IPD game is played between any pair of neighboring agents regardless of their string types. In the other setting, it is played only between neighboring agents with the same string type. Using these two settings, we examine the effect of the IPD game between agents with different representation schemes on strategy evolution. We also examine the effect of the number of different representation schemes in a population (i.e., the number of species) on strategy evolution.
Keywords :
game theory; iterative methods; agent population; agent string types; game playing; game strategy encoding; game strategy evolution; iterated prisoner´s dilemma game; neighborhood structure; representation schemes; spatial IPD game; two-dimensional grid world; Australia; Computational intelligence; Computer science; Encoding; Games; Standards; Thin film transistors; Iterated prisoner´s dilemma games; encoding; evolution of game strategies; evolutionary games; representation schemes; spatial IPD games;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Evolutionary Computation (CEC), 2012 IEEE Congress on
Conference_Location :
Brisbane, QLD
Print_ISBN :
978-1-4673-1510-4
Electronic_ISBN :
978-1-4673-1508-1
Type :
conf
DOI :
10.1109/CEC.2012.6256442
Filename :
6256442
Link To Document :
بازگشت