Information Technology Reference
In-Depth Information
Genetic algorithms are an important class of evolutionary algorithms, where the
evolutive mechanism is realized by two steps: i) a recombination step where two
(or more) individual solutions are mixed in some way for producing new solutions,
and ii) a mutation step, where some individuals are casually changed in a certain
small percentage. These two ingredients reflect the two main forces used by nature
in genetic evolution. Of course, the manners and the extent of these two evolutive
steps greatly influence the kind of results, and greatly depend on the context of
application of the genetic algorithms we consider. Moreover, the representation of
solutions determines many different forms of recombinations and mutations. Usu-
ally solutions can be represented by strings; therefore recombination corresponds to
the ways of exchanging substrings, while mutations are casual replacement of single
symbols (a certain number of them).
Memetic algorithms are a generalization of genetic algorithms, where the more
abstract notion of meme replaces the notion of gene. Many possible interpretations
of this concept, very often informally defined, have been developed. However, gen-
erally, a meme has a recombinant character and a mobility greater than those exhib-
ited by a gene. In fact, a meme can be spread around in many copies and in many
places, and can take part in mutation and recombinations with other memes where
the number of copies, and the mutation and recombination extent are context de-
pendent. For example, the number and the sizes of components it exchanges and/or
the number of partners with which it interacts result from the “place” where the
meme lives. The metaphor underlying memes is that of cultural units. They spread
and evolve with the interactions among social groups which come in contact, by
exchanging ideas, habits, words, mental attitudes, and so on. The composition of
these groups constantly changes at a very quick rate. Formally, these features can
be expressed in many ways, but their overall effect is to make the evolutive, selec-
tive, and mutational aspects more articulated and complex than the corresponding
genetic mechanisms. This fact relates to the complexity and fluidity of cultural evo-
lution with respect to the genetic evolution. A possible way of formalizing memetic
mechanisms is provided by membrane structures, where we can assume genetic
algorithms operating on internal membranes (genetic pools), and additional rules
exchanging individuals and even sub-membranes, among different membranes, and
where the levels of nested membranes could represent different levels of recombi-
nation.
Immunological algorithms are another type of evolutionary algorithms. Their
specificity is based on the notions of clonal selection and negative selection ,which
are inspired by the corresponding theories developed since the mid 20th century
in immunology. Clonal selection is realized when selection is performed by some
recognition mechanism. For example, only solutions which match, to a certain de-
gree, some target elements are produced in many copies and this confers to them
more chance of becoming chosen in the further evolution process (this is the kind
of selection B cells undergo by means of antibody-antigen matching). Negative
selection occurs when the target elements which solutions have to satisfy constitute
an unknown space , viewed as a set complement of some known space (this is the
Search WWH ::




Custom Search