Particle Swarm Optimization Based On Dynamic NW Small World Network

This paper has proposed two improved particle swarm optimizations based on NW small world network(NWPSO) and virus propagation model(VPPSO), which can solve the problem of conferencing to local optimal caused by the reduction of population density. The neighborhood typology of NWPSO algorithm is initialized by k-nearest neighbor network, and increases its small world characteristic by adding the random edge during the iteration. VPPSO takes the share information as visual propagating in the k-neighbor network, and control information transmission rate by adjusting the transmit probability. The proposed NWPSO algorithm and VPPSO algorithm both have the capacity to adjust the population diversity. What’s more, the NW neighborhood topology of NWPSO algorithm can divide the particle swam into community structures, which can gain the population diversity more efficiently.The simulation results show that, NWPSO and VPPSO have more effective optimization than some classical PSO algorithms in the single-mode state functions, and can avoid local optimal during the optimization of multi-modal functions and rotated multi-modal functions.