Аннотация
As the research field called Genetic Programming has
shown during the last decade, it is possible not only
to write computer programs by hand but also to let the
computer itself develop programs that solve given
problems. This is achieved by simulating natural
evolution on the computer for "breeding" programs
that are well adapted to a specific problem
environment. The use of mechanisms found in nature can
lead to solutions to complex problems that by far
outperform any man-made approaches. The reasons are
that complex problems often are difficult to solve
analytically and many other possible approaches are not
accessible to the human way of thinking. The use of the
mechanisms of evolution based on genetic variation and
"survival of the fittest" is only one example.
Another example are Artificial Neural Networks that
imitate clusters of nervous cells and their
interactions for solving difficult problems (inspired
among others by the human brain).
The here presented work explores a different and new
approach to adopting problem solving methods found in
nature. It uses the natural cell control mechanism
called Gene Regulation that according to modern
molecular genetics is the basis of the cooperation
between and differentiation into all the different
cells in living creatures. The most astonishing example
of self-organization between simple units that
cooperate to solve complex problems is not the
interaction between nervous cells on the basis of
mutual electrical activation through explicit and
directed connections. It is the interaction between all
kinds of cells in a living creature which is based on
the diffusion of messages in the form of produced
substances. This interaction is much more powerful and
flexible than the neural interaction because of many
reasons. The main reason is, that a cell in this
context is not only a simple unit which can have
different levels of activation, but it is a complex
system with many behavioural possibilities. The
communication between the cells not only bases on
different activation intensities but on many different
message types which (also depending their intensity)
can have very sophisticated effects on the behaviour of
a cell.
This new programming and control paradigm has been
combined with genetic programming for breeding
"multicellular" programs (which probably is the
only feasible way of producing them). The system that
implements this combination can not only be used to
create programs with a new modular structure which has
several advantages. It also is a great tool for
developing systems of cooperating autonomous units like
Amorphous Computers and Multiagent Systems.
Пользователи данного ресурса
Пожалуйста,
войдите в систему, чтобы принять участие в дискуссии (добавить собственные рецензию, или комментарий)