Landscape produced by genetic programming where Constriction
hinders
Particle Swarm Optimisation
Swarm of ten Particles without Constriction
10 particles without constriction factor.
In this run 7 generations were needed to find an optima.
Note the Swarm remains energetic and easily moves away from initial position
to one of the optimal areas.
Swarm of ten Particles with Constriction factor of 0.7
A Swarm of 10 particles operating on the same landscape as above
from the same initial conditions,
except now we use a constriction factor.
The swarm loses energy so fast that, despite the gradient,
it never reaches either optimal area.
Instead it remains near 5.83,6.04 forever.
Example
where constriction helps.
Presented at
Swarm Intelligence Symposium 2005.
Back to main XPS page.
W.B.Langdon
4 June 2005