Evolutionary Algorithm

The most suitable evolutionary algorithms to solve optimization problems in continuous domains are evolutionary strategies [Sch81,Rec73], genetic algorithms [Hol75,Gol89a] with real coding [Gol91] and evolutionary programming [FOW66,Fog95]. For evaluating CIXL2 we have chosen real coded genetic algorithms, because they are search algorithms of general purpose where the crossover operator plays a central role. The general structure of the genetic algorithm is shown in Figure 4.

Figure 4: Structure of the genetic algorithm, $t$ is the current generation.

Nevertheless, CIXL2 could be applied to any evolutionary algorithms with a crossover or similar operator. On the other hand, the real codification is the most natural one in continuous domains, each gene representing a variable of the function. In this way, the precision of the solution only depends on the data type used to store the variables.

Our objective is the comparison of the behavior of the proposed crossover against other crossovers. This comparison must be made in a common evolutionary framework that is defined by the features of the genetic algorithm. For the definition of such features, we have taken into account the previous studies on the matter. In the following paragraphs we will describe in depth the different components of our genetic algorithm.