Information Technology Reference
In-Depth Information
x
2
^
E
(e)
2
1
0.2
0
−1
0.1
−2
x
1
−3
e
0
−2
0
2
4
−2
−1
0
1
2
3
(a) Epochs=0
0.22
x
2
^
E
(e)
2
1
0.2
0
−1
0.18
−2
x
1
−3
e
0.16
−2
0
2
4
−2
−1
0
1
2
(b) Epochs=20
0.5
x
2
^
2
E
(e)
0.4
1
0
0.3
−1
−2
0.2
x
1
−3
e
0.1
−2
0
2
4
−4
−2
0
2
Error Rate (Test) = 0.230
2
0.8
H
^
Error Rate
S
1.8
0.6
1.6
0.4
1.4
0.2
epochs
epochs
0
0
10
20
30
40
50
0
10
20
30
40
50
(c) Epochs=50
Fig. 3.3 The Gaussian two-class dataset of Example 3.3 at different epochs of the
MEE linear discriminant. The upper three graphs at the left side show the dataset
with the linear decision border (solid line); the graphs at the right side show the
error PDF. The bottom graphs show the evolution of the Shannon entropy and the
error rate for the training set (solid) and test set (dotted).
