Geology Reference
In-Depth Information
Gene Manipulation
Genetic Algorithm
6. Integrate individuals created based on
gene manipulation and created based on
traditional genetic procedure to make child
population for next generation.
Using a flexible implicit redundant representation
encoding method, the number of control devices
on a specific floor and the sensor locations are
specified. To enhance the search performance
and reduce the simulation time, new genetic al-
gorithm is proposed. Gene manipulation genetic
algorithm (GMGA) (Cha 2008) uses engineering
judgment to create encoding variables to search
non-dominated individuals based on novel re-
combination/mutation mechanism. The novel
mechanism uses average, maximum, minimum,
and random perturbation values of the variables
of each two adjacent non-dominated individuals
of current Pareto-optimal front. To define the
percentage of individuals in current population
that undergo gene manipulation process, GMGA
uses a gene manipulation ratio (GMR) which
ranges from 0.1 to 0.3 in research investigated
(Cha 2008). The other remaining population is
filled with individuals that undergo traditional
genetic operating process such as crossover and
mutation. The gene manipulation procedures of
the GMGA are (Cha et al. 2009; Cha et al. 2011a):
The number of new individuals to create in
each section of the Pareto-optimal front can be
calculated (Cha 2009):
Number of new string (i) =
round
Each distance (i)
Total distance
(10)
×
Pop. Size
×
GMR
An example that identifies how the gene
manipulation process is working to create new
individuals in sections between non-dominated
individuals is shown in Figure 7. The GMR is 0.1
and population size is 100 and then 10 new indi-
viduals will be generated using GMGA. The two
adjacent individuals are selected from the current
Pareto-optimal front. Based on Euclidean distance
between the individuals in Pareto-optimal front,
the numbers of new individuals for each section
are determined using Equation (10). Several new
gene instances are created using one of four gene
manipulation mechanisms to generate a new indi-
vidual in a specific section as shown in Figure 7.
New gene instance variable are generated using
four operations (Cha 2008) as stated in Equations
(11)-(14) (Cha et al. 2011a):
1. Non-dominated sorting in the current popu-
lation using any MOGA selection method
such as NSGA series or SPEA series
2. Define GMR to determine the number of
individuals that undergo gene manipulation
process and determine exact number of each
number that is created between individuals
in Pareto front (see Equation 10)
3. Select representative individuals in each
identified section of the Prato front to be
used for new individuals created
4. Generate new gene instances using one of
four gene manipulation mechanisms.
5. Insert created gene instances in the repre-
sentative individuals
for s=1:
t
{
if
k
<= number of new individuals created in each
section,
s
(i.e. create the first new individual)
for
v
=1:
n{
1
j
∑
P s k v
( ,
, )
P v i
( , )
=
(11)
n
j
c
i
=
1
Search WWH ::
Custom Search