Training of the Artificial Neural Networks using States of Matter Search Algorithm

• Published in: International Journal of Intelligent Systems and Applications in Engineering Vol. 8; no. 3; pp. 131 - 136
• Main Author: Gulcu, Saban
• Format: Journal Article
• Language: English
• Published: Selçuk Üniversitesi 28.09.2020
• Subjects: Astronomi & Uzay; Bilgisayar Bilimleri; Bilim-Teknoloji; Mühendislik; Türkiye; Konya; training of artificial neural networks; feed-forward artificial neural networks; states of matter search algorithm; optimization; Artificial neural networks
• ISSN: 2147-6799; 2147-6799
• DOI: 10.18201/ijisae.2020363532

In recent years, technology has been developing very rapidly in the field of artificial intelligence. In this development, Artificial neural networks (ANNs) have taken a huge place. The human brain has an excellent understanding structure. The brain makes this understanding through neuron cells. ANN aims to solve some complex problems by establishing the perception structure of human over neurons in the computer environment. A multilayer perceptron (MLP) is a class of artificial neural networks. MLP has the ability to learn using inputs and expected outputs. In order to do this, weight values in MLP are constantly updated according to the inputs and expected outputs. Thus, weight values are tried to be kept at an optimum level. Therefore, this problem is an optimization problem. In this study, the State of Matter Search meta-heuristic algorithm was used to optimize the weight values in MLP, called SMS-MLP. In the experiments, five classification datasets (xor, balloon, iris, breast cancer, heart) were used. The SMS-MLP algorithm was compared with the previous six algorithms (GWO-MLP, ACO-MLP, GA-MLP, PBIL-MLP, PSO-MLP and ES-MLP) in the literature. The experimental study shows that the SMS-MLP algorithm is more efficient than the other six algorithms.