Deficiencies of Best-chromosome-wins Dominance in Evolutionary Optimization of Stationary Functions
[ 1 ] Instytut Informatyki, Wydział Informatyki i Telekomunikacji, Politechnika Poznańska | [ P ] pracownik | [ S ] student
2024
rozdział w monografii naukowej / referat
angielski
- evolutionary algorithms
- diploidy
- polyploidy
- dominance
EN In evolutionary computation, diploid genotypes (i.e., genotypes with two chromosomes) are traditionally used mostly in the context of optimization of non-stationary problems. Recent research, however, suggested that the use of diploid genotypes with mechanisms such as best-chromosome-wins can improve the performance of evolutionary algorithms even for stationary problems. In this paper we test the effectiveness of diploidy and polyploidy (i.e., genotypes with more than two chromosomes) with best-chromosome-wins on mathematical benchmarks. We verify the effect and the importance of the crossover operator on the behavior of evolutionary algorithms with diploidy and polyploidy. We explore the inner workings of evolutionary algorithms with diploidy and polyploidy in order to better understand their performance. We find that the results reported in previous papers on the best-chromosome-wins dominance for stationary functions may have been overly optimistic, and the use of diploidy with best-chromosome-wins does not enhance the search process for such functions.
01.08.2024
471 - 474
20
140