A similarity-based surrogate model for enhanced performance in genetic algorithms

In spite of their ability to deal with difficult optimization problems, genetic algorithms, in general require a large number of evaluations in order to find an optimal or a satisfactory near-optimal solution. When expensive simulations are involved in the optimization process, using genetic algorithms as optimization tools can become unattractive. The use of surrogate models is an interesting alternative to overcome that drawback, either by replacing expensive evaluations or allowing for exploration of the search space. In this paper we introduce a surrogate model based on similarity measures into genetic algorithms in order to enhance the performance in optimization problems, under a fixed budget of simulations. Numerical experiments are conducted in order to assess the applicability and the performance in constrained and unconstrained optimization problems. The results show that the present framework arises as an attractive alternative to improve the final solutions with a fixed budget of expensive evaluations.