Swarm Intelligence

The first of these versions is the traditional particle swarm. An lbest

sociometry was defined, with each neighborhood comprising the indi-

vidual and its two adjacent neighbors. For version 1, then,

r

p

i

and

r

p

g

were

simply the best points found by the individual and by its best-

performing neighbor (including itself).

In version 2, the individual’s previous best performance was stereo-

typed, that is, particle i was attracted toward its cluster’s average previous

best performance as well as the performance of the best neighbor.

Version 3, replacing

r

p

g

with g’s (the best neighbor’s) cluster center, at-

tracted particle i toward the center of g’s cluster,