
best. As the effect of these terms is to adjust the velocity of the particle,
the effect on its trajectory is not to move it with the gradient, but to
cause the particle to oscillate around the previous best point—sometimes
with the gradient and sometimes against it. The gradient then is used
simply to keep the particle oriented in the region of previous successes,
similar to methods in EC, and is not followed in the traditional way of
hill climbers.
A particle moves in a stochastic oscillatory trajectory through the
problem space, sampling around relatively optimal local points, while an
evolutionary individual searches by changing position through ...