6.7 baCkpropagatioN Neural Network (bpNN) model 243
Calculate the outputs of the output layer:2.
net
5
= W
35
× H
1
+ W
45
× H
2
- q
5
= (1)(0.269) + (1)(0.269) - 1 = -0.462
Y = 1/[1 + exp(- net
5
)] = 1/[1 + exp(0.462)] = 0.386
d
5
= Y(1 - Y )(T - Y ) = 0.386 (1 - 0.386)(0 - 0.386) = - 0.0915
d
3
= H
1
(1 - H
1
) × ∑W
3j
d
j
- 0.269(1 - 0.269)[(1)(-0.0915)] = -0.0180
d
4
= H
2
(1 - H
2
) × ∑W
4j
d
j
= 0.269(1 - 0.269)[(1)(-0.0915)] = -0.0180
Calculate the 3. ∆w of the output layer:
∆W
35
= hd
5
H
1
= (10)(-0.0915)(0.269) = -0.246
∆W
45
= hd
5
H
2
= (10)(-0.0915)(0.269) = -0.246
∆q
5
= -hd
5
= -(10)(-0.0915) = 0.915
Calculate the 4. ∆w of the hidden layer:
∆W
13
= hd
3
x
1
= (10)(-0.0180)(-1) = 0.180
∆W
23
= hd
3
x
2
= (10)(-0.0180)(-1) = 0.180
∆q
3
= -hd
3
= - (10)(-0.0180) = 0.180 ...