The role of different genetic operators in the optimization of magnetic models

摘要

In the paper we continue the analysis of the influence of genetic operators on the efficiency of global optimization of Ising model. For such a model the physical problem of finding the minimum energy is defined. Its solution enables to determine the phase state of system in the given temperature and consequently the phase transition point. The process of minimum energy search is NP-hard. We showed that for a selected set of crossover operators we can obtain similar behavior but different exact numerical characteristics. Here we study, how the other procedures and parameters can influence the optimization process. We take into account such features like: selection, mutation, objective function or niching. We try also to build a hybrid algorithm with a specific problem-related local search procedure. We hope that the presented results can indicate the great possibilities of magnetic models as a test problem.