Effectiveness of new Multiple-PSO based Membrane Optimization Algorithms on CEC 2014 benchmarks and Iris classification

Abstract

This paper proposes a new Multiple-PSO based Membrane Algorithm (M_PSO_MA), which is a one-layered parallel and distributed membrane structure with seven different membranes, i.e., a skin membrane containing six elementary membranes using update rules of different PSOs. The idea is inspired from Frankenstein’s PSO (Oca et al. in IEEE Trans Evol Comput 13:1120–1132, 2009) which shows that by integrating components in novel ways effective optimizers can be designed. M_PSO_MA, is designed using the variants of PSO as given in Oca et al. (2009), which although are not hybridized but are applied simultaneously on the same population set. Each elementary membrane sends the best solution and the corresponding particle to the skin membrane, where the best solution is determined among all the collected solutions, which happens to be the best solution of the complete swarm. This global best solution and the corresponding particle are then communicated to all other membranes by the skin membrane besides retaining a copy for itself. The proposed algorithm, M_PSO_MA provides good quality solutions, but may sometime require slightly additional computational time. Therefore, its modified variant namely, MM_PSO_MA is also presented with an objective to minimize the computational time. Both, the proposed variants are compared with the existing best membrane algorithm determined in Singh et al. (Appl Math Comput 246:546–560, 2014) on the basis of CEC-2014 benchmark optimization problem set. In the second part of our paper M_PSO_MA and MM_PSO_MA are applied to solve a real life problem of Iris classification and the results are compared with Mutated PSO, namely MO-MPSO (Wang et al. in Int J Innov Comput Inform Control 9:2963–2977, 2013). The extensive results prove the supremacy of M_PSO_MA over other PSO based Membrane Algorithms. It is concluded that the proposed algorithm can be effectively used for solving complex real life problems as well.