An experimental analysis of the effects of migration in parallel genetic algorithms

The paper presents some experimental results concerning parallel genetic algorithms. Genetic algorithms are a well-established technique for the solution of large optimization problems; a parallel version has been proposed for them, based on the concept of migration. Several parameters concerning migration deeply affect the performance of the approach, but it is often difficult to optimize their value in order to obtain the best result. The paper presents a system which produces good solutions to the traveling salesman problem using parallel genetic algorithms, and reports some results concerning the influence of the parameters on the performance of the system