Information Technology Reference
In-Depth Information
Chapter 13
Evolutionary Computation
13.1 Introduction
Evolutionary computation is a computing system which applies natural evolution
and adaptive thought(Yao,2006). Darwin's evolutionism is a robust mechanism
for searching and optimization, which has great impacts on the development of
computer science, in particular artificial intelligence. The majority of organisms
evolve by natural selection and sexual reproduction. Natural selection determines
which individuals in population can survive and reproduce, while sexual
reproduction is the guarantee of gene mixture and recombination in the next
generation. The principle of natural selection is "to select the superior and
eliminate the inferior, survival of the fittest".
Early in the 1960s, the above features of natural evolution had drawn
Holland's great interest. In those years, he and his students were studying on how
to establish a machine learning system. He found that machine learning can be
realized not only by an individual's adaptation, but also by plurivoltine
evolutions of a population. By inspiration of Darwin's evolution thought, Holland
realized that in order to obtain an excellent learning algorithm, a reproduction of
a population with multiple candidate strategies, instead of only one strategy,
should be built and improved. As this idea originated from genetic evolution,
Holland called this research field genetic algorithm. Genetic algorithm was not
well known until his currently famous monograph "Adaptation in Natural and
Artificial Systems "(Holland, 1975) was published in 1975. His monograph
systematically introduces the basic theory of genetic algorithm, and forms a
foundation for genetic algorithm. In the same year, De Jong finished his doctoral
dissertation “An Analysis of the Behavior of a Class of Genetic Adaptive
Systems”. In this doctoral dissertation, De Jong applied Holland's schema theory
467
