SEARCH, Computational Processes in Evolution, and Preliminary Development of the Gene Expression Messy Genetic Algorithm
Hillol Kargupta
Electronic mail address: hillol@eecs.wsu.edu.
Department of Computer Science,
Washington State University,
Pullman, WA 99164-2752
Abstract
This paper considers the issue of scalable search with little domain knowledge and explores implications in the context of evolutionary computation. It presents the Search Envisioned As Relation and Class Hierarchizing (SEARCH) framework introduced elsewhere [26, 31] for developing a theoretical understanding of the issue and argues that scalable evolutionary search needs efficient techniques for detecting relations among the members of the evolutionary search space. It offers a perspective of this argument in the context of natural gene expression (representation transformations that construct the protein from the DNA). It also reports on the preliminary development of the gene expression messy genetic algorithm (GEMGA) [27, 28] that exploits the understandings developed here. Theoretical claims are also substantiated by experimental results for a test bed, comprised of different large, multimodal, scaled problems.
