A Unified Paradigm for Parallel Genetic Algorithms

A. Kapsalis, G. D. Smith, V. J. Rayward-Smith

Research output: Chapter in Book/Report/Conference proceedingChapter

5 Citations (Scopus)


Genetic Algorithms contain natural parallelism. There are two main approaches in parallelising GAs. The first one is parallelising individual functional components of a standard, sequential GA. The only difference with the sequential GA is in the computation speed. The second approach more closely resembles the real life simultaneous evolution of species, which is the central theme in GAs. Algorithms following this approach are still referred to as GAs but are different from Holland's standard GA. For these algorithms it is not the improvement in computation speed that is the driving factor, but the efficiency with which they search a given solution space. We describe a number of the most common parallel GA methods found in the literature and mention practical issues concerning their implementation in a Transputer based system. We go on to introduce the Unified Parallel GA system, based on our GA toolkit, GAmeter, which allows the user to select one or more of the GA methods described, by setting various parameters. Finally we present results for the Steiner tree Problem in Graphs (SPG).
Original languageEnglish
Title of host publicationEvolutionary Computing
EditorsTerence Fogarty
Place of PublicationHeidelberg
Number of pages19
Publication statusPublished - 1994

Publication series

NameLecture Notes in Computer Science

Cite this