Information Technology Reference
In-Depth Information
(3) Quality function serving as environment;
(4) Selection process for generating offspring;
(5) Genetic operators, such as mutation, recombination;
(6) Parameters for controlling algorithm itself.
In natural world, selection operation in individual's environment is
accomplished locally. Natural selection equation which is abstracted from
environment may lead to the risk of repetition. Parallel genetic algorithm has
been successfully applied to solve many combinatorial problems. Each species is
simulated on a processor. A parallel genetic algorithm can be described as
follows:
1. Given
N
individuals with different initial phenotype;
2. Calculate local maximum value of each individual;
3. Selection: select near-neighbour individuals for mating;
4. Create offspring by recombination and mutation;
5. Goto step 2.
There are two differences between parallel genetic algorithm and classical
genetic algorithm. One is that individual mating is locally finished; the other is
that individual utilizes hill climbing method.
Therefore, parallel genetic algorithm utilizes genotype learning and
phenotype learning. Further, each individual survives in a certain environment,
and crossover and selection are accomplished in such environment, such as 2-D
grid. Individual's neighbor does not connect to the grid of genotype and
phenotype space. Parallel genetic algorithm originates mainly from classical
genetic algorithm. In genetic algorithm, individual is regarded as the result of
selecting "the best behavior". On the contrary, parallel genetic algorithm is based
on internal driver. Parallel genetic algorithm has more differences over general
genetic algorithm. Its evolution is based on driver of internal self-Organization.
13.8 Classifier System Boole
In 1987, Wilson developed a classifier system for solving boolean problem:
Boole(Wilson, 1987). A boolean function is a mapping from strings with length
of L to {0,1}. Learning function means that the ability to correctly map any
input(a string) to 0 or 1 is obtained.
