Biomedical Engineering Reference
In-Depth Information
Figure 8. Mutation operation
ANN's Population
1
2
3
4
5
ANN's Population
1
2
3
4
5
Mutation
Mutation
Sort
Sort
Fitness
Fitness
N-1
N
N-1
N
Figure 9. Crossover operation
ANN's Population
ANN's Population
1
2
3
4
5
1
2
3
4
5
1
2
3
4
5
Crossover
Crossover
Sort
Sort
N-1
N
N-1
N
N-1
N
Fitness
Fitness
high limit of weights and gradients, the type of
activation functions (lineal, threshold, hyperbolic
tangent and sigmoid). To the RANN the param-
eters codified are: continuous or epochs training
and the number of internal iterations to evaluate
an input. In the Figure 11 we can see all the pa-
rameters for the ANN architecture.
Results
It is necessary to use Recurrent ANN architectures
for the prediction of time series and for modelling
this type of problems. To the training and valida-
tion of this new RANN architecture, 2 classical
time series in the field of the statistics have been
used (William, 1990) (Box, 1976), the number
of sunspots and the tobacco production in USA.
 
Search WWH ::




Custom Search