Information Technology Reference
In-Depth Information
Evaluation of
Initial Population
Random Initial
Population
Selection
(Tournament)
Replacement
Evaluation
Mutation
Recombination
Fig. 9.18
Evolutionary loop of a basic GA
evaluated using a problem-speci
c objective function called
fitness. Individuals are
selected based on their
fitness (using a series of tournaments), these selected
individuals are called Parents. These parents are used to generate new individuals
using two basic genetic (search) operations, recombination (random recombination
of two or more individuals), and mutation (random modi
cation of a single indi-
vidual). These newly generated individuals are called Offspring, since they share
(genetic) similarities with the Parents used to generate them. Finally, the best
individuals (among Parents and Offspring) are selected and replace the initial
population. The algorithm is iterated until a stop criterion is reached; for instance,
when all individuals are identical (convergence of the algorithm) or after a pre-
speci
ed number of iterations.
9.6.2 Algorithmic Choices
In this work, the genome is composed of 16 variables: the
first, an integer ranging
from 1 to 58, characterizes the number of the electrode selected, the 15 others are
binary and correspond to the inclusion (or not) of each frequency to compute the
slope criterion. An example of a genome is given in Fig.
9.19
. Each genome de
nes
the electrode and the frequencies on which to perform the regression as illustrated in
Fig.
9.20
.
9.6.2.1 Genetic Operators
The main search operators used with the GA are crossover (recombination) and
mutation; both are described in detail next.
Electrode number
Binary part
15
0
0
0
0
1
1
0
1
1
0
1
0
1
1
1
Fig. 9.19
Example of a genome in the GA
