A Fast Particle Swarm Optimization Algorithm by Refining the Global Best Solution

A Fast Particle Swarm Optimization (FPSO) is proposed to improve the convergence response speed for some potential application scenarios such as the online or dynamical optimization environment which requires the fast convergence ability of an optimizer. Classical gradient-based optimization methods are good at finding the local optimal value of a convex region yet usually failure in searching the global optimal value of a multimodal problem. To further develop the characteristics of PSO with respect to the fast convergence and the global optimization, a pseudo-gradient method is proposed for calculating the approximate gradient at the location of the global best solution (gBest) of a swarm to refine the convergence accuracy of the gBest so as to accelerate the local convergence speed. The experimental results show that the performance of the proposed algorithm is significantly better than those of the five chosen competitive algorithms on a series of benchmark test functions with different characteristics. Furthermore, the sensitivity of the new introduced parameter in the proposed algorithm is empirically analyzed by a special experiment for recommending its best range of value.