Biomedical Engineering Reference
In-Depth Information
classification. This process can be continued as
long as the classification results are equal, or at
least similar, to those obtained using the overall
set of variables. Therefore, the GA determines
how many and which wave numbers will be
considered for the classification.
In this approach, each individual in the genetic
population is described by n genes, each represent-
ing one variable. With the binary encoding each
gene might be either 0 or 1, indicating whether the
gene is active or not and, therefore, if the variable
should be considered for classification.
The evaluation function guides the pruning
process in order to get individuals with a low
number of variables. To achieve this, the function
should help those individuals that, besides clas-
sifying accurately, make use of fewer variables.
In this particular case the function increments
the percentage of active genes to the MSE (me-
dium square error) obtained by the ANN, so the
individuals with less active genes - and with a
similar classification performance - will have a
lower fitness and consequently better chances to
survive.
Table 4 shows the results of several runs of
pruned search. It should be borne in mind that
each solution was provided in a different execu-
tion and that, within the same execution, only
Table 4. Classification with pruned search: percent accuracy
Low Concentrations
Low Concentrations
Training (134)
Training (134)
Validation (39)
Validation (39)
Comercial (2)
Comercial (2)
Run1: Selected Variables
[42 77]
Run1: Selected Variables
[42 77]
129 (96.27%)
129 (96.27%)
23 (58.97%)
23 (58.97%)
1 (50%)
1 (50%)
Run2: Selected Variables
[52 141]
Run2: Selected Variables
[52 141]
115 (85.82%)
115 (85.82%)
22 (56.41%)
22 (56.41%)
0 (0%)
0 (0%)
Run3: Selected Variables
[102 129]
Run3: Selected Variables
[102 129]
124 (92.54%)
124 (92.54%)
25 (64.10%)
25 (64.10%)
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: 0.0001 stop criterion: mse=5 or epochs=500.000
learning rate: 0.0001 stop criterion: mse=5 or epochs=500.000
High Concentrations
High Concentrations
86
86
43
43
21
21
Run1: Selected Variables
[42 77]
Run1: Selected Variables
[42 77]
83 (96.51%)
83 (96.51%)
34 (79.07%)
34 (79.07%)
21 (100%)
21 (100%)
Run2: Selected Variables
[52 141]
Run2: Selected Variables
[52 141]
82 (95,35%)
82 (95,35%)
35 (81.40%)
35 (81.40%)
20 (95.24%)
20 (95.24%)
Run3: Selected Variables
[102 129]
Run3: Selected Variables
[102 129]
83 (96.51%)
83 (96.51%)
33 (76.74%)
33 (76.74%)
21 (100%)
21 (100%)
ANN Configuration
Topology: 2 / 10 / 60 / 5 learning rate: 0.0001 stop criterion: mse=1 or epochs=500.000
ANN Configuration
Topology: 2 / 10 / 60 / 5 learning rate: 0.0001 stop criterion: mse=1 or epochs=500.000
Search WWH ::




Custom Search