Abstract:To improve the global performance of particle swarm optimization (PSO), a novel cooperative particle swarm optimizer(PSO) with two swarms in different behavior is proposed. In the method, two swarms are utilized to search the best solution. Particles in the first swarm are attracted to the position which determined by the best position of current generation and that the particles derived so far. Particles in the second swarm learn from his or her own and other individuals' mistakes, their positions and velocities are updated according to the worst states which the particles derived so far and of the current generation. The best and the worst solutions are exchanged between the two swarms, information will flow mutually between the two swarms when the best and the worst pariticles are not in the same swarm. To indicate the effectiveness of the proposed algorithm, several benchmark functions and forward ANN with three layers are performed in simulation experiments, the results show that the given method has good performance with comparing to other PSOs. Copyright © 2010 Binary Information Press.

A cooperative approach to particle swarm based on two swarms with different behavior