Yapay Zeka Hızlı Bakış & Çıkarımlar
AI InsightDeneysel çalışma sonucunda BWO algoritmasının arama sürecindeki çeşitlilik eksikliği sebebiyle ortaya çıkabilecek olan erken yakınsama probleminin iyileştiği gözlemlenmiştir.
📌 3 Maddede Bu Makale
Özet
Özet 1 : Bu çalışmada son zamanlarda literatüre sunulmuş bir meta-sezgisel optimizasyon algoritması olan Beyaz balina optimizasyon (Beluga whale optimization, BWO) algoritmasının problemlere daha uygun sonuçlar üretmesi amacıyla iyileştirilmiş bir versiyonu geliştirilmiştir. Beyaz balinaların yüzme, avlanma ve ölme özellikleri modellenerek geliştirilmiş olan BWO algoritmasında yer alan arama süreçlerinde uygunluk uzaklık dengesi (fitness-distance balance, FDB) seçim yöntemi uygulanmıştır. BWO algoritması ve FDBBWO ismi verilerek geliştirilen algoritmanın performanslarını test etmek için CEC2020 test fonksiyonları kullanılmıştır. Bu test fonksiyonları üzerinde 30, 50 ve 100 boyut için algoritmalar test edilmiştir. Elde edilen test sonuçlarına Friedman analizi yapılarak algoritmaların performans sıraları belirlenmiştir. Ayrıca Wilcoxon sıralı işaret testiyle de sonuçlar üzerinde anlamlı derecede farklılıklar oluşup oluşmadığı incelenmiştir. Deneysel çalışma sonucunda BWO algoritmasının arama sürecindeki çeşitlilik eksikliği sebebiyle ortaya çıkabilecek olan erken yakınsama probleminin iyileştiği gözlemlenmiştir. Bu sayede yerel optimum noktalara takılma olasılığı azaltılmıştır. Ayrıca geliştirilen algoritma literatüre son zamanlarda sunulmuş olan 3 farklı algoritmayla karşılaştırılmıştır. Karşılaştırma sonuçlarına göre FDBBWO, diğer meta-sezgisel algoritmalara göre daha üstün bir performans sergilemektedir. Özet 2 : In this study, an improved version of the Beluga whale optimization (BWO) algorithm, which is a meta-heuristic optimization algorithm recently presented in the literature, is developed to provide better solutions for the problems. The fitness-distance balance (FDB) selection method was applied in the search processes in the BWO algorithm, which was developed by modeling the swimming, preying and falling characteristics of beluga whales. CEC2020 benchmark functions were used to test the performance of the BWO algorithm and the algorithm named FDBBWO. The algorithms were tested on these test functions for 30, 50 and 100 dimensions. Friedman analysis was performed on the test results and the performance ranks of the algorithms were determined. In addition, Wilcoxon rank sum test was used to analyze whether there were significant differences in the results. As a result of the experimental study, it is observed that the BWO algorithm improves the early convergence problem that may arise due to the lack of diversity in the search process. In this way, the possibility of getting stuck at local optimum points is reduced. In addition, the developed algorithm is compared with 3 different algorithms that have been recently presented in the literature. According to the comparison results, FDBBWO has a superior performance compared to other meta-heuristic algorithms.
Abstract
In this study, an improved version of the Beluga whale optimization (BWO) algorithm, which is a meta-heuristic optimization algorithm recently presented in the literature, is developed to provide better solutions for the problems. The fitness-distance balance (FDB) selection method was applied in the search processes in the BWO algorithm, which was developed by modeling the swimming, preying and falling characteristics of beluga whales. CEC2020 benchmark functions were used to test the performance of the BWO algorithm and the algorithm named FDBBWO. The algorithms were tested on these test functions for 30, 50 and 100 dimensions. Friedman analysis was performed on the test results and the performance ranks of the algorithms were determined. In addition, Wilcoxon rank sum test was used to analyze whether there were significant differences in the results. As a result of the experimental study, it is observed that the BWO algorithm improves the early convergence problem that may arise due to the lack of diversity in the search process. In this way, the possibility of getting stuck at local optimum points is reduced. In addition, the developed algorithm is compared with 3 different algorithms that have been recently presented in the literature. According to the comparison results, FDBBWO has a superior performance compared to other meta-heuristic algorithms.