Computer science and information processes

Training an artificial neural network using the PSOJaya hybrid optimization algorithm

E. M. Kazakova

Institute of Applied Mathematics and Automation – branch of Kabardino-Balkarian Scientific Center of the Russian Academy of Sciences, 360000, Russia, Nalchik, 89 A Shortanov street

Abstract: Particle swarm optimization (PSO) and Jaya are heuristic optimization algorithms that are used to find optimal solutions in optimization problems. Each of these methods has its own advantages and disadvantages, and the choice between them depends on the specific optimization problem and performance requirements. This paper proposes a hybrid of PSO and Jaya algorithms to improve optimization efficiency. In this paper PSO, Jaya, PSOJaya are used as artificial neural network (ANN) training methods for the classification task on the Balance Scale dataset. The results of the hybrid algorithm are compared with the results of the Backpropagation, PSO and Jaya algorithms. The test calculations compare the algorithms based on the mean, median, standard deviation, and "best" minimum error value after 30 simulations. The experiment results show that the ANN trained with PSOJaya has higher accuracy than the one trained with Backpropagation, PSO and Jaya.

Keywords: heuristic algorithm, optimization, particle swarm method (PSO), Jaya, Backpropagation, hybrid algorithm, pipeline hybridization, ANN, classification

UDC: 004.023

MSC: 68T07

Received: 01.11.2023
Revised: 04.11.2023
Accepted: 14.11.2023

DOI: 10.35330/1991-6639-2023-6-116-95-102

Bibliographic databases:

•