Biomedical Engineering Reference
In-Depth Information
Figure 12. Evolution of the GA in the first 200 generations (10 internal iterations in the hyperbolic
ANN)
0 .0 5
0 .0 4 5
0 .0 4
No time decreased activation
0 .0 3 5
0 .0 3
0.01
0 .0 2 5
0 .0 2
0 .0 1 5
0 .0 1
0.3
0.6
0.1
0 .0 0 5
0
1
1 1
2 1
3 1
4 1
5 1
6 1
7 1
8 1
9 1
1 0 1
1 1 1
1 2 1
1 3 1
1 4 1
1 5 1
1 6 1
1 7 1
1 8 1
1 9 1
Table 4 shows the configuration of the ANN
parameters in order to test the system and the
simulated time decreased activation with 5 in-
ternal iterations in the ANN. We also stop the
genetic algorithm after 200 generations.
As we can see in the Figure 15, in this case,
a time decreased activation of 0.1 shows the best
evolution for 5 internal iterations in the ANN.
In this case (as the hyperbolic activation func-
tion tests) the RANN needs more speed to make
the simulation of the action potential between
cycles.
In these tests we can see the hyperbolic acti-
vation function with time decreased of 0.3 and
10 internal iterations in the ANN shows the best
evolution. Then we make another test without
limitation in the number of generations of the GA
and with different number of hidden neurons.
As we can see in the Table 5, with 3 neurons
in the hidden layer we can obtain the lower MSE,
and the best solution for the prediction of this
time series.
Results for Tobacco Production in
USA Time Series
With the same operation of the previous point
and with different values for the simulated time
decreased activation, in the Table 6 we can observe
the results for different combinations.
As we can see in Table 7, all the MSE are very
similar, but the hyperbolic activation function
with only one hidden neuron produces the best
value. In the following Figures we can see the
structure of this RANN and the comparative of
the predictions that it produces.
Search WWH ::
Custom Search