Information Technology Reference
In-Depth Information
10
8
6
4
2
0
−2
−4
−6
−8
−10
−10
−8
−6
−4
−2
0
2
4
6
8
10
x
Fig. 12.13
Approximation of the test function of Eq. (
12.25
)(
dashed line
), using an RBF neural
network with Gaussian basis functions
1.5
1
0.5
0
−0.5
−1
−1.5
10
5
10
5
0
y
0
−5
x
−5
−10
−10
Fig. 12.14
Test function of Eq. (
12.26
)
of the test function obtained by a one hidden layer FNN with sigmoidal basis
functions is shown in Fig.
12.16
. Finally the approximation achieved by RBF
neural network with Gaussian activation functions is depicted in Fig.
12.17
.
5. Test function 5 given in Eq. (
12.27
) over the domain D D Œ10;10 Œ10;10
(Fig.
12.18
):
f.x;y/D 1:9e
.0:02x
2
0:02y
2
/
tanh.0:3y/
(12.27)
The approximation results for the 2D function of Eq. (
12.27
), obtained by an
NN with Hermite basis functions, are depicted in Fig.
12.19
. The approximation