Política de migração para Metaheurísticas Híbridas usando Modelo Paralelo de Ilhas

Abstract

Optimization problems are present in several areas of knowledge and with the advancement of society their complexities increase. In the last decades several evolutionary population algorithms have been proposed to solve optimization problems. Each of these algorithms has its own characteristics that influence in a different way the evolutionary process and the quality of the optimal solution. The Island Model is a strategy of parallelization of evolutionary population algorithms. In the Island Model the set of candidate solutions is divided into subsets known as islands. Each island evolves its set of solutions through its own algorithm in parallel with the others. Periodically, the islands exchange solutions among themselves through the migration process. The movement of solutions between islands is conditioned to the topology of the model and a set of rules that compose the migration policy. This work proposes a new alternative of hybrid implementation for the Island Model, inspired by the natural phenomenon of stigmergy, called Stigmergy Island Model. The connections of Stigmergy Island Model have weights adjusted on each migration and that influence in choice of destination island for each migrant solution. The main objective with the Stigmergy Island Model is to propose a self-organizing structure for solving optimization problems through distinct evolutionary population algorithms. One of the advantages of the Stigmergy Island Model is that the user can apply several algorithms in its topology and the migration policy itself will identify the best configuration for the topology and distribution of the set of solutions among islands to solve the problem.