Different versions of particle swarm optimization for magnetic problems

The paper presents the application of Particle Swarm Optimization into the magnetic problems where the structure of sample, its stoichiometry and the character of magnetic interactions is described by some well known models. We use three different models or approximations what enables to use three different versions of PSO: binary, real-number and discrete (multi-state). We show that, in order to prepare the efficient code leading to the correct results, we have to include some changes. The most important is the modification of the relative strength of the cognitive and social factors determining the value of velocity. We show also that the computational hardness of the optimization problem depends on the choice of physical parameters. This feature makes it possible to use the presented cases as an interesting testing tool. We compare also our results with the results obtained by using genetic algorithms found either in references or generated by our own code.