Résumé: The genetic algorithm (GA) and simulated annealing algorithm (SA) are empirically compared for the problem of optimizing the topological design of a network. In addition to the usual problem of optimizing only the placement of links, in this letter the number and placement of concentrators are also decision variables for a class of problems using a real set of concentrators, links, and traffic. The average GA solution cost less than the average SA solution.

Langue: Anglais