Information Technology Reference
In-Depth Information
sqrt
sqrt
a
a
b
b
a
b
b
a
/
a
b
Figure 1.10.
Tree mutation in genetic programming. The arrow indicates the
mutation point. Note that the mutation operator creates a randomly generated
branch at the mutation point.
are randomly permuted (Figure 1.11). In this case the overall shape of the
tree remains unchanged.
In summary, in GP the operators resemble more of a conscious mathema-
tician than the blind way of nature. But in adaptive systems the blind way of
nature is much more efficient and systems such as GP are highly limited in
evolutionary terms. For instance, the implementation of other operators in
GP, such as the simple yet high-performing point mutation (Ferreira 2002a),
is unproductive as most mutations would have resulted in syntactically in-
correct structures (Figure 1.12). Obviously, the implementation of other
a
a
/
/
sqrt
sqrt
a
a
b
b
b
b
Figure 1.11.
Permutation in genetic programming. The arrow indicates the point
of permutation. Note that the arguments of the permuted function changed places
in the daughter tree. After permutation the shape of the tree remains unchanged.
