Simultaneous Perturbation Particle Swarm Optimization

In this paper, we propose two hybrid algorithms of the particle swarm optimization and the simultaneous perturbation optimization method. The proposed algorithms can utilize local information of an objective function and global shape of the function at the same time. The first information is given by the simultaneous perturbation. The second one is from the particle swarm optimization. The proposed scheme has good properties of global search and efficient local search. However, the algorithms themselves are very simple and easy to implement. Moreover, this method only requires values of the function similar to the original particle swarm optimization and the simultaneous perturbation method. Three examples including an application for a neural network are shown.