Variations and Comparisons

This chapter explores implementations of the particle swarm paradigm, shows some variations of the algorithm, discusses whether the particle swarm is an evolutionary algorithm, and compares the performance of various versions of particle swarms for optimization. Approaches include limiting the maximum allowed particle velocity, including a constriction coefficient and an inertia weight. It considers a new way to look at evolution, focusing on the role of self-organization. The chapter also examines crossover and mutation processes, and population topology as they relate to particle swarms. Moreover , this chapter compares versions of the particle swarm paradigm that include experiments in which selection is added to the particle swarm paradigm, comparisons of the inertia weight, and constriction factor approaches, and a test in which particle swarms are initialized asymmetrically with respect to the global optimum. The effectiveness of the particle swarm algorithm comes from the interactions of particles with their neighbors. The sociometry of the particle population interacts significantly with function. Evolutionary and social algorithms have much in common and their fusion seems inevitable. Some evolutionary operations are analogous to particle swarm methods, but the differences between approaches are considerable.