Training of artificial neural networks with the multi-population based artifical bee colony algorithm


Kirankaya C., Gorkemli Aykut L.

NETWORK-COMPUTATION IN NEURAL SYSTEMS, cilt.33, sa.1-2, ss.124-142, 2022 (SCI-Expanded) identifier identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 33 Sayı: 1-2
  • Basım Tarihi: 2022
  • Doi Numarası: 10.1080/0954898x.2022.2062472
  • Dergi Adı: NETWORK-COMPUTATION IN NEURAL SYSTEMS
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Applied Science & Technology Source, Computer & Applied Sciences, EMBASE, MEDLINE, zbMATH
  • Sayfa Sayıları: ss.124-142
  • Anahtar Kelimeler: Artificial neural networks, metaheuristic algorithms, artificial bee colony algorithm, multi-population-based strategy, PARTICLE SWARM OPTIMIZATION, HYBRID, PERFORMANCE
  • Erciyes Üniversitesi Adresli: Evet

Özet

Nowadays, artificial intelligence has gained recognition in every aspect of life. Artificial neural networks, one of the most efficient artificial intelligence techniques, is remarkably successful in computers' acquisition of the learning and interpretation capabilities of humans and attainment of meaningful results. Whether artificial intelligence networks can yield meaningful results is directly related to how the network is trained. The traditional algorithms, which are used to train artificial intelligence networks, do not always yield successful results in complicated problems and real-life problems. Metaheuristic algorithms are efficient techniques developed in order to figure out time-consuming and challenging problems fast and as optimally as possible. This study makes use of the artificial bee colony algorithm, which has been widely used recently in solving many kinds of problems so as to train artificial neural networks efficiently. Within this study, different strategies are used on subpopulations, so that the algorithm can search without getting tangled with the local optima. And also same and different parameter settings are considered for each population to reflect different search behaviours. The proposed approach was analysed through applied results of different data sets. The results yielded that the used strategies can be promising alternatives to the current search algorithms.