Genetic Algorithms and Multimodal Search - Advancing Artificial Intelligence Through Biological Process Applications - page 237

The Populations: Secondary
Population

The Genetic Operators: Crossover

As it was pointed before, the crossover operator
recombines the genetic material of individuals of
both populations. This recombination involves a
random individual from secondary population
and a representative of the genetic pool.
This representative will represent a potential
solution offered by the genetic pool. As a unique
individual can not verify this requirement, the
representative will be formed by a subset of genes
of different individuals on the genetic pool. Gath-
ering information from different partial solutions
will result in producing a valid global solution.
Therefore, the value for every gene of the
representative will be randomly chosen among all
the individuals in the genetic pool. After a value
is assigned to all the genes, this new individual
represents not a partial, unlike every one of the
individuals separately, but a global solution.
Now, the crossover operator will be applied.
This crossover function will keep the secondary
population diversity, so the offspring will contain

The individuals of the secondary population are
quite different from the previous. In this case, the
genes of the individuals of the secondary popula-
tion can take any value throughout the whole space
of possible solutions, a fact that makes them able
to offer global solutions to the problem. This is
not possible in genetic pool because their genes
are restricted to different sub-ranges.
The evolution of the individuals at the genetic
pool will be carried out by traditional GA rules.
The main difference lies in the operator crossover.
In this case a modified crossover will be used.
As the information is stored in isolated popula-
tions, the parents who produce the new offspring
will not belong to the same population. Hence,
the genetic pool and secondary population are
combined instead. In this way information of
both populations will be merged to produce the
most fitted offspring.

Figure 4. Crossover operator for hybrid two-population genetic algorithm

Representative
(First Parent)

…

G 11 G 12 G 13 G iN

…

…

G 21 G 22 G 23 G iN

Uniform
Crossover

…

G P1 G P2 G P3 G PN

…

…

G iN

G 11 G 12 G 13

Second Parent

…

G 21 G iN

G 22 G 23

…

G P1 G P2 G P3 G PN