Modeling Dependencies of Loci with String Classification According to Fitness Differences

Miwako Tsuji, Masaharu Munetomo, and Kiyoshi Akama

Abstract. Genetic Algorithms perform crossovers effectively when we can identify a set of loci tightly linked to form a building block. Several methods have been proposed to detect such linkage. Linkage identification methods investigate fitness differences by perturbations of gene values and EDAs estimate the distribution of promising strings. In this paper, we propose a novel approach combining both of them, which detects dependencies of loci by estimating the distribution of strings classified according to fitness differences. The proposed algorithm called the Dependency Detection for Distribution Derived from df (DDDDD or D⁵) can detect dependencies of a problem which is difficult for EDAs requiring lower computation cost than linkage identifications.

LNCS 3103, p. 246 ff.

Full article in PDF

lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2004