Computational Processes in Evolution and the Gene Expression Messy Genetic Algorithm

This paper makes an eeort to project the theoretical lessons of the SEARCH (Search Envisioned As Relation and Class Hierarchizing) framework introduced elsewhere (Kargupta, 1995; Kargupta & Goldberg, 1996) in the context of natural evolution (Kargupta, 1996c) and introduce the gene expression messy genetic algorithm (Kargupta, 1996a; Kargupta, 1996b) (GEMGA)|a new generation of messy GAs that directly search for relations among the members of the search space. The GEMGA is an O(jj k (` + k)) sample complexity algorithm for the class of order-k delin-eable problems (Kargupta, 1995) (problems that can be solved by considering no higher than order-k relations) in sequence representation of lengthànd alphabet set. Unlike the traditional evolutionary search algorithms, the GEMGA emphasizes the computational role of gene expression and uses a transcription operator to detect appropriate relations. Theoretical conclusions are also substantiated by experimental results for a test bed, comprised of diierent large, multimodal, scaled problems.