A Dynamic Island-Based Genetic Algorithms Framework

This work presents a dynamic island model framework for helping the resolution of combinatorial optimization problems with evolutionary algorithms. In this framework, the possible migrations among islands are represented by a complete graph. The migrations probabilities associated to each edge are dynamically updated with respect to the last migrations impact. This new framework is tested on the well-known 0/1 Knapsack problem and MAX-SAT problem. Good results are obtained and several properties of this framework are studied