A comparison between evolutionary-algorithms parallelization in CUDA architecture

Evolutionary algorithms try to calculate the best solutions for the complex problems. These algorithms inspire from the existed processes in nature and evolutionary principles. They calculate the optimum solution by iterative processes and searching problem space to find the optimal or near optimal solution. To improve convergence probability to optimal solution and finding the optimum solutions with high precision and accuracy, they utilize a large number of iteration. However, some of the evolutionary algorithms have parallel nature because they search for the population members, separately. Thus, the speed of these algorithms can be increased and consequently the effect of initial population size and repetitions on execution time can be decreased. This paper argues parallelizing on CUDA architecture and compare the speed evolutionary algorithms including particles swarm optimization(PSO), firefly algorithm (FA), bat algorithm (BA) and artificial fish swarm algorithm (AFSA). More clearly, we compare evolutionary algorithms in centralized and parallelized approaches. Evaluations are done by some simulations and results show the various aspects of parallelizing on CUDA framework.