A Model for Parallel Operators in Genetic Algorithms

In this chapter we analyze a model for applying parallel operators in Genetic Algorithms. Here we focus on crossover and varying mutation applied separately in parallel, emphasizing the gains on performance that can be achieved from the concurrent application of operators with different and complementary roles. We analyze the model in single population Genetic Algorithms using deterministic, adaptive, and self-adaptive mutation rate controls and test its performance on a broad range of classes of 0/1 multiple knapsack problems. We compare the proposed model with the conventional one, where varying mutation is also applied after crossover, presenting evidence that varying mutation parallel to crossover gives an efficient framework to achieve higher performance. We also show that the model is superior for online adaptation of parameters and contend that it is a better option for co-adaptation of parameters. In addition, we study the performance of parallel operators within distributed Genetic Algorithms showing that the inclusion of varying mutation parallel to crossover can increase considerably convergence reliability and robustness of the algorithm, reducing substantially communication costs due to migration.