Information Technology Reference
In-Depth Information
is called the phenotype or body, that is, the entity that faces the environment
and does all the work. A simple example of a replicator/phenotype system is
the DNA/protein system of life on Earth. It is believed that for life to move
beyond a very rudimentary stage, the phenotype threshold must be crossed
(e.g., Dawkins 1995; Maynard Smith and Szathmáry 1995).
Similarly, the entities of both GAs and GP (simple replicators) survive by
virtue of their own properties. Understandingly, there has been an effort in
the last years in the scientific community to cross the phenotype threshold in
evolutionary computation. The most outstanding effort is developmental
genetic programming or DGP (Banzhaf 1994) where binary strings are used
to encode mathematical expressions. The expressions are decoded using a
five-bit binary code, called genetic code. However, contrary to its analogous
natural genetic code, this “genetic code”, when applied to binary strings,
frequently produces invalid expressions (in nature there is no such thing as a
structurally incorrect protein). Therefore, a huge amount of computational
resources goes into editing these illegal structures, which limits this system
considerably. Not surprisingly, the gain in performance of DGP over GP is
minimal (Banzhaf 1994; Keller and Banzhaf 1996).
Gene expression programming is an example of a full-fledged replicator/
phenotype system where the chromosomes/expression trees form a truly func-
tional, indivisible whole (Ferreira 2001). Indeed, in GEP there is no such
thing as an invalid expression tree or program. Obviously, the interplay of
GEP chromosomes and expression trees requires an unambiguous transla-
tion system to transfer the language of chromosomes into the language of
expression trees. Furthermore, we will see that the structural organization of
GEP chromosomes allows the unconstrained modification of the genome,
creating the perfect conditions for evolution to occur. Indeed, the varied set
of genetic operators developed to introduce genetic modification in GEP
populations always produce valid expression trees, making GEP a simple
artificial life system, well established beyond the replicator threshold.
To help the non-biologist understand the fundamental difference between
gene expression programming and the other genetic algorithms and why GEP
is such a leap forward in evolutionary computation, it is useful to know a
little more about the structure and function of the main players of biological
gene expression and how they work together. What follows is a very brief
introduction to the structure and function of the main molecules of informa-
tion metabolism and how mutation in proteins relates to evolution. If you
wish to pursue these questions further, any textbook on biochemistry will
