Biomedical Engineering Reference
In-Depth Information
principal components that can describe the data
set: two variables.
As the amount of wave numbers remains
constant in the course of the selection process,
this approach defines the fitness of each genetic
individual as the mean square error reached by
the ANN at the end of the training process.
Table 5 shows the results of several runs of
pruned search. It should be highlighted that dif-
ferent runs provide only one valid solution each
time.
Since the genetic individuals contain only
two genes and the crossover operator has to use
the unique available crossing point (between the
genes), only half of the information from each
parent would be transmitted to its offspring ,
This converts the fixed search approach into a
random search when only two genes constitute
the chromosome.
Hybrid Two-Population Genetic
Algorithm
As discussed previously, the main disadvantage
of non-multimodal approaches is that they discard
optimal solutions while a final or global solution
is preferred. But there are situations where the
final model is extracted after analysing different
solutions of the same problem.
Table 6. Classification with hybrid two-population genetic algorithm
Training (134)
Training (134)
Validation (39)
Validation (39)
Comercial (2)
Comercial (2)
Low Concentrations
Low Concentrations
Run1: Selected Variables
[89 102]
Run1: Selected Variables
[89 102]
127 (95.77%)
127 (95.77%)
29 (74.36%)
29 (74.36%)
0 (0%)
0 (0%)
Run1: Selected Variables
[87 102]
Run1: Selected Variables
[87 102]
130 (97.01%)
130 (97.01%)
28 (71.79%)
28 (71.79%)
0 (0%)
0 (0%)
Run1: Selected Variables
[88 89]
Run1: Selected Variables
[88 89]
120 (89.55%)
120 (89.55%)
29 (74.36%)
29 (74.36%)
0 (0%)
0 (0%)
ANN Configuration for low concentrations
Topology: 2/ 10 / 60 / 7
ANN Configuration for low concentrations
Topology: 2/ 10 / 60 / 7
learning rate: 0.001 stop criterion: mse=2 or epochs=500.000
learning rate: 0.001 stop criterion: mse=2 or epochs=500.000
High Concentrations
High Concentrations
86
86
43
43
21
21
Run1: Selected Variables
[89 102]
Run1: Selected Variables
[89 102]
83 (96.51%)
83 (96.51%)
35 (81.39%)
35 (81.39%)
21 (100%)
21 (100%)
Run1: Selected Variables
[87 102]
Run1: Selected Variables
[87 102]
83 (96.51%)
83 (96.51%)
36 (83.72%)
36 (83.72%)
21 (100%)
21 (100%)
Run1: Selected Variables
[88 89]
Run1: Selected Variables
[88 89]
82 (95.35%)
82 (95.35%)
39 (90.69%)
39 (90.69%)
21 (100%)
21 (100%)
ANN Configuration
Topology: 2 / 10 / 60 / 5 learning rate: 0.001 stop criterion: mse=2 or epochs=500.000
ANN Configuration
Topology: 2 / 10 / 60 / 5 learning rate: 0.001 stop criterion: mse=2 or epochs=500.000
Search WWH ::
Custom Search