A hierarchical subpopulation particle swarm optimization algorithm

Based on the metaphor of specialization and cooperation in hierarchical social organization, a new particle swarm optimization (PSO) algorithm, hierarchical subpopulation PSO (HS-PSO), was proposed. In HS-PSO, the entire population is divided into several subpopulations which are arranged in a hierarchy. The subpopulations at the same level of the hierarchy evolve relatively independently and cooperate with each other via their respective best particles. The particles at different levels are assigned special tasks and thus different parameters are employed for them for a good balance of exploration and exploitation. Two versions of HS-PSO which use the same or different kinds of PSO algorithms for the particles at different levels were presented. The efficiency of HS-PSO was verified by comparing it with some variants of PSO in the optimization of 5 benchmark functions.