Information Technology Reference
In-Depth Information
shown in Figure 12.1, this population is midway to the peak with a success
rate of 48%. Such systems, although adaptively healthy, are not very effi-
cient and are called
healthy but weak
.
As shown in Figure 12.1, the success rate increases abruptly with muta-
tion rate until it reaches a plateau around
p
m
= 0.022. In terms of dynamics,
this is reflected in a more pronounced oscillatory pattern in average fitness
and an increase in the gap between average and best fitness. For obvious
reasons, populations evolving with maximum performance are called
healthy
and strong
. The next three dynamics (plots
c
,
d
, and
e
) are all drawn from the
performance plateau or peak. Note, however, that from the peak onward, an
increase in mutation rate results in a decrease in performance until populations
are totally incapable of adaptation (see Figure 12.1). In these populations,
the gap between average and best fitness continues to increase (see plots
f
through
i
) until the plot for average fitness reaches the bottom and populations
become totally random and incapable of adaptation (see plot
i
). Note also
that some populations, despite evolving under excessive mutation rates, are
still very efficient. For instance, the evolutionary dynamics shown in plot
f
,
p
m
= 0.046 and
Ps
= 93%, is extremely efficient. This kind of population is
called
unhealthy but strong
(with “unhealthy” indicating the excessive mu-
tation rate). The next population,
p
m
= 0.15 and
Ps
= 43% (plot
g
), is not as
efficient as the previous one and therefore is called
unhealthy and weak
.
The last two dynamics were obtained for populations outside the finger
region of the mutation plot. In the first,
p
m
= 0.45 and
Ps
= 3% (plot
h
), the
plot for average fitness is not far from the minimum position obtained for
totally random populations (plot
i
). Thus, populations of this kind are called,
respectively,
almost random
and
totally random
. Note that oscillations on
average fitness are less pronounced in these plots. It is worth noticing that
totally random populations (the presence of elitism at
p
m
= 1.0 is, in this
experiment with 500 individuals, insignificant in terms of average fitness)
are unable to find a perfect solution to the problem at hand. And this tells us
that every time a perfect solution was found, a powerful search mechanism
was responsible for this and not a random search.
Transposition
The evolutionary dynamics of transposition are shown in Figure 12.3. As you
can see, the dynamics obtained for both RIS and IS transposition are similar to
the ones obtained for mutation. And this is worth pointing out, because not all
operators display this kind of dynamics (see the dynamics of recombination
