Impact of the quality of random numbers generators on the performance of particle swarm optimization

Intelligent search algorithms are highly efficient to solve problems when it is not possible to use exaustive search. Particle Swarm Optimization (PSO) is a bio-inspired technique to perform search in continuous and hyperdimensional spaces. Despite it is common used to solve real world problems, a deeper study on the impact of the quality of Random Number generators has not been made yet. In this paper, we compare the performance of four variations of PSO algorithms in several benchmark functions considering five different Random Number Generators. PSO with inertia and constricted were analyzed. Global and local topologies were explored as well. The five different Random Numbers Generators are derived from Linear Congruential Generator (LCG) and the Marsaglia's algorithm. We showed that PSO algorithms need random number generators with a minimum quality. However, we also showed that no significative improvements were achieved when we compared high quality random number generators to medium quality Random Number Generators.