
Artificial Neural Networks ◾ 213
11.
a.
T =
− +
− −
+ −
1
3
0 2 2
2 0 2
2 2 0
b. We have to test for two conditions: (1) that the weights T
ii
are
equal to 0; and (2) that the weights satisfy Hebb’s rule (i.e., for-
mula (4.10)). Verify that T satises both of these conditions.
c. We will demonstrate this using matrix multiplication notation.
Tvv =
− +
− −
+ −
+
−
+
1
3
0 2 2
2 0 2
2 2 0
1
1
1
=
+
−
+
1
3
4
4
4
Applying the sign() function yields:
sign( )Tvv vv=
+
−
+
=
1
1
1
which means that v is stable. e calculation for the other vector is
similar.
15. A possible solution is to use all the available data for training, leav-
ing out only a single example to be u