An Investigation on the Effect of Migration Strategy on Parallel GA-Based Shortest Path Routing Algorithm

Abstract

Genetic algorithm (GA) is one of the alternative approaches for solving the shortest path routing problem. In previous work, we have developed a coarse-grained parallel GA-based shortest path routing algorithm. With parallel GA, there is a GA operator called migration, where a chromosome is taken from one sub-population to replace a chromosome in another sub-population. Which chromosome to be taken and replaced is subjected to the migration strategy used. There are four different migration strategies that can be employed: best replace worst, best replace random, random replace worst, and random replace random. In this paper, we are going to evaluate the effect of different migration strategies on the parallel GA-based routing algorithm that has been developed in the previous work. Theoretically, the migration strategy best replace worst should perform better than the other strategies. However, result from simulation shows that even though the migration strategy best replace worst performs better most of the time, there are situations when one of the other strategies can perform just as well, or sometimes better.

A Parallel Genetic Algorithm for Configuring Defect Detection Methods

“…En la literatura hay diseñados muchos métodos de trabajo para este tipo de operadores [18] [19], pero en este caso se ha optado por el método que suele proporcionar los mejores resultados en la mayoría de los problemas. En este caso, los peores individuos de cada isla son sustituidos por los mejores individuos de las demás islas cada k generaciones de la isla.…”