An Improved Social Learning Particle Swarm Optimization Algorithm with Selected Learning

Particle Learning Optimization (PSO) is a novel heuristic algorithm that has undergone decades of evolution. Social learning particle swarm optimization (SL-PSO) proposed by Cheng, Jin et al. in 2016 [1] remarkably improves the PSO algorithm by applying multi-swarm learning strategy. Nevertheless, randomness on setting inertia and choosing learning objects gives rise to an unbalanced emphasis on global search, and thus impairs convergence rate and exploitation ability. The proposed ISL-PSO algorithm strengthens global search capability through modelling selected learning mechanism, in which learning objects are selected through generated learning possibility subjected to Gauss distribution. Furthermore, ISL-PSO algorithm models condition-based attraction process, in which particles are attracted to the center by calculating transformed distance between particles and the center. By applying the strategies, ISL-PSO improves convergence speed and accuracy of the original algorithm.