Hybrid PSO-Jaya Algorithm for Solving Various Optimization Problems

Abstract

This article proposes a hybrid algorithm PSO-Jaya, based on two heuristic algorithms — PSO (Particle swarm optimization) and Jaya. The main idea of the PSO-Jaya hybrid algorithm is to use PSO for global research of the solutions space and when PSO ceases to improve the results, which indicates a possible hit in the local optimum, Jaya is connected for the best solution throughout space, taking into account the best solutions already found. This allows the hybrid algorithm to combine efficient exploration of large areas of the solution space with the ability to minimize the probability of hitting local optimum. Two problems are used to evaluate the effectiveness of the PSO-Jaya hybrid algorithm: functions optimization and training artificial neural network for classification task Glass Identification. In calculation tests, the PSO, Jaya, PSO-Jaya algorithms are compared in view of their mean, median, standard deviation and "best" minimum error. In this connection, 50 simulations-based test functions and 30 network simulations were fulfilled. The performance analysis the PSO and Jaya algorithms was carried out as well as the hybrid algorithm for test-based problems. In all test cases, the PSO-Jaya algorithm achieved the best performance in terms of convergence speed and ability to avoid local optimum.