In this paper, we re-visit the parallel PSO schemes and adopt a different perspective to parallelize PSO. In the adopted scheme, the original particle swarm is divided into several sub-swarms under the condition that the total particle numbers are the same, thus keeping nearly the same computational tasks. Each slave processor then individually runs a part of the sub-swarms. Experiments are also conducted to show the property and the performance of the scheme.
Relation:
Network-Based Information Systems (NBiS), 2015 18th International Conference, pp.616-619