Approches évolutionnaires pour le problème de partitionnement de graphes

Abstract

We have studied through this thesis the application of evolutionary approaches, in particular the genetic algorithm (GAs), for solving the graph partitioning problem. Known for their simplicity and efficiency, GAs have been widely used to solve complex problems. Most of the research works has focused on designing new genetic operators or modifying the operating mechanism of the GA. Although this improved the efficiency of GAs, the performances of this latter is strongly related to the genetic representation of the solutions. From this perspective, we presented a state of the art of genetic representations with an a throughout study of their properties and characteristics in order to propose new representations based on the advantages of those present in the literature. The empirical study we conducted on a meticulously chosen data set from the literature shows the superiority of the proposed encoding schemes. These letters are based on the p-median problem that uses both vertex and edge information