THE SPECIATION IN GENETIC ALGORITHMS FOR PRESERVING POPULATION DIVERSITY AND OPTIMIZATION OF FUNCTIONS WITH SUBOPTIMAL SOLUTIONS

Abstract

Background. Genetic algorithms are used in various tasks and show advantages compared to other optimization methods, which may not always be applicable. However, there are cases when they can’t provide the best possible solution. One of them is the premature convergence of the population to a single solution. As population diversity decreases, the search domain becomes limited, and potential solutions may be overlooked. This scenario is particularly common in multimodal functions where multiple local optima exist. To address this, a class of techniques known as niching techniques has been developed. These methods preserve population diversity and prevent premature convergence to suboptimal solutions.

Materials and Methods. In this work we investigate the method of speciation and how it helps to find solutions for given tasks. For this, several experiments were conducted, in which the highest value of the function was found in the given interval. It was compared how results of the optimization differed if we used speciation and didn't. To assess the diversity of speciation, the values of the average fitness of the population and the standard deviation of the values of the individuals in the population were compared. We also evaluated how speciation helps with optimization for tasks with suboptimal solutions, comparing how many successful solutions were obtained in experiments with and without speciation.

Results and Discussion. The results show that the speciation method preserves population diversity and improves optimization outcomes for multimodal functions. In the experiments where speciation was applied, the population maintained a higher level of diversity, as indicated by a larger standard deviation in population individuals' values. It resulted in increasing of the number of successful solutions in tasks with multiple local optima.

Conclusion. Speciation effectively preserves population diversity and helps to avoid premature convergence in genetic algorithms. This leads to better optimization results, especially in tasks with multiple local optima. This highlights the importance of diversity-preserving techniques, such as speciation, in addressing the limitations of genetic algorithms, especially in complex optimization tasks.

Keywords: genetic algorithms, optimization, niching techniques, speciation