in this experiment the particles seek out the region of most red
in this case the region at the bottom left is the reddest with the top right section being a large local maximum it's easy for the swarm to get caught up in
ticks represent the personal best of a particle. lines represent where the particle is currently in relation to their best
it's possible that the third swarm may get stuck in the sub optimal local maxima of the top right hand corner. how could we mitigate this problem?
| these particles are just moving in straight lines | these particles are also moving in straight lines but have an attraction to return to where they have their personal best | these particles move in straight lines, with an attraction to their personal best, but also an attraction to the personal best of two fixed other particles |
conclusion: to be honest a straight hill climbing algorithm could probably out preform these.
i think a big part of it is that this problem has a pretty small search space and the particle swarm stuff won't really shine until we get into higher dimension problems...