Biomedical Engineering Reference
In-Depth Information
computing, an area that is very intimately con-
nected to all signs of life: evolution. After a general
discussion of algorithms derived from evolution
( evolutionary algorithms or evolutionary computing ),
we consider in more detail the most modern
branch of this area, genetic programming .
would need to be able to learn in a fashion similar
to children—a very clear indication of him taking
inspiration from biology. “Now the learning pro-
cess may be regarded as a search [...]. Since there
is probably a large number of satisfactory solu-
tions, the random method seems to be better than
the systematic. It should be noticed that it is used
in the analogous process of evolution” [8, p. 459].
So, already in 1950 several ideas were voiced
that would lead the way to evolutionary algo-
rithms. The notion of a soft kind of randomness,
which would later become mutation and crosso-
ver, the notion of intelligent behavior as the goal
of these algorithms, the notion of a search process
to achieve learning and problem solving, and the
notion of kinship to evolution in Nature were all
entertained in this early article by Alan Turing.
In 1962, a budding computer scientist 2 from the
University of Michigan published a paper entitled
“Outline for a Logical Theory of Adaptive Sys-
tems” in which he proposed most of what later
became known as genetic algorithms [9] . Holland
wrote: “The study of adaptation involves the study
of both the adaptive system and its environment,”
thus foreshadowing the necessity to define a fit-
ness function as a stand-in for the environment.
He then proposed to look at the adaptive system
as a “population of programs” and emphasized
the advantage of looking at adaptation from the
viewpoint of a population: “There is in fact a gain
in generality if the generation procedure operates
in parallel fashion, producing sets or populations
of programs at each moment rather than individu-
als” [9, p. 298] . Here Holland correctly identified
the strength that populations of solutions bring to
a problem when applied and tested in parallel.
“The generated population of programs will act
upon a population of problems (the environment)
in an attempt to produce solutions. For adaptation
to take place the adaptive system must at least be
able to compare generation procedures as to their
efficiency in producing solutions.” What Holland
17.2 HISTORY AND VARIANTS OF
EV OLUTIONARY COMPUTIN G
Evolutionary algorithms or evolutionary com-
puting is an area of computer science that applies
heuristic search principles inspired by natural
evolution to a variety of different domains, nota-
bly to parameter optimization or other types of
problem solving traditionally considered in arti-
ficial intelligence.
Early ideas in this field developed at a time
when computers were barely commercially sold.
Alan Turing was one of the first authors to cor-
rectly identify the power of evolution for the pur-
pose of solving problems and exhibiting intelligent
behavior. In his 1950 essay entitled “Computing
Machinery and Intelligence,” [8] Turing consid-
ered the question of whether machines could
think. He was concerned that digital computers,
despite their power and universality, would only
be capable of executing programs deterministi-
cally. He felt that this would not be sufficient to
produce intelligent behavior. “Intelligent behav-
ior presumably consists in a departure from the
completely disciplined behavior involved in com-
putation, but a rather slight one, which does not
give rise to random behavior” [8] , p. 457. Compu-
tation here refers to the only known form of digital
computation at the time: deterministic computa-
tion. Turing pointed out that a digital computer
with a random element would be an interesting
variant of such a machine, especially “when we
are searching for a solution of some problem.” It
was clear to Turing that machines at some point
2 Computer science as a discipline did not even exist then. It is only in hindsight that we call it that; officially, it
was called communication science at the time.
 
Search WWH ::




Custom Search