Biomedical Engineering Reference
In-Depth Information
Figure 5. Pseudo code for nutation: Delta initialization
IF (not Bi)
Gi = Gi + Ii
IF (Gi > LIM_SUP_GEN)
Gi = LIM_INF_GEN
Ii = Ii - Delta
ENDIF
ENDIF
(
LIM
_
SUP
_
IND
)
(
LIM
_
INF
_
IND
)
Delta =
IND
_
POOL
values from the genetic pool. Therefore the GA
would be able to maintain multiple solutions in
the same population. The crossover operator does
not change the genetic pool because the last one
only acts as an engine to keep the diversity. This
process is summarised in Figure 4.
Genetic Algorithm with Division into
Species
Another proposed solution is an adaptation of the
niching technique. This adaptation consists on the
division of the genetic population into different and
independent subspecies. In this case the criterion
that determines the specie for a specific individual
is done according to genotype similarities (simi-
lar genotypes will form isolated species). This
classical concept has been provided with some
improvements in order to, not only decrease the
number of iterations needed for obtaining solu-
tions, but also increase the number of solutions
kept within the genetic population.
Throughout evolution the individuals tend to
gather into different species; as every one these
species will be adapted to an environment and
will evolve differently, there will be the same
number of optimal individuals than of species.
Sometimes the individuals also split up from the
initial group due to different reasons and they
undergo crossover in other place with individuals
of other groups.
Broadly speaking, the grouping technique
into species divides the initial population into
groups—the species—where the individuals have
similar characteristics. In this way, each group
will be specialised in a given area of search space;
each species will be responsible for searching
the solutions that exist in its area, which are also
different from other solutions provided by other
species. The evolution of every species will be
The Genetic Operators: Mutation
The mutation operator increases the value of in-
dividual genes in the genetic pool. It introduces
new information into the genetic pool for the
representative to use it and finally, by means of
the crossover operator, introduces it in secondary
population.
It should be noted that the new value will have
upper limit, so when it is reached the new gene
value will be reset to the lower value.
When generations advance the increment
amount is reduced, so the increment applied to the
individuals in the genetic pool will take lower val-
ues. The different increments between iterations
are calculated bearing in mind the lower value for
a gene (LIM_INF_IND), the upper value for that
gene (LIM_SUP_IND) and the total number of
individuals in the genetic pool (IND_POOL), as
Figure 5 summarises. In such way, first generations
will explore the search space briefly (coarse-grain
search) and, as the search process advance, it is
intended to do a more exhaustive route through
all the values that a given gene may have (fine-
grain search).
Search WWH ::




Custom Search