This applet shows two particle swarm optimisers attempting to find the minimum of a 2-dimensional function called Schaffer's F6. A cross-section of the function is shown below (not an easy function!). The minimum is at the origin.
Particle swarm optimisers are so named because the candidate solutions appear like swarms of particles that "fly" through the space.
Notice that one population explores more rapidly but fails to converge. The second learns more slowly, but ultimately converges to a better solution.
This is an example of the classical trade-off in population-based optimisation methods between exploration and exploitation.
Copyright Cara MacNish, CSSE, UWA
Images and animations copyright their respected creators.