Parallel genetic algorithms take advantage of concurrent execution to obtain better results
and better use of the machine’s hardware. Usually there are multiple subpopulations that
evolve concurrently and communicate through a defined migration policy, to achieve better
exploration of the search space, for example. Non conventional genetic operators are the
ones inspired by some natural organisms, such as viruses and bacteria, to modify the
genetic algorithm architechture. It’s common that these operators use auxiliar populations
containing special individuals to obtain better genetic variability. This work proposes an
implementation of a parallel genetic algorithm that makes use of the recombination by
bacterial transformation genetic operator, and then compares its performance with both
sequential genetic algorithms that make use of this same operator and parallel versions that
use conventional genetic operators. The results show that the presented implementation
in general brought a higher speed of convergence, higher robustness, and precision, if
compared to the other implementations that are used.
|
