SYS735 Intelligent Control Systems KaCC
2-4 FFNN Matlab NNT examples 1 27-Jan-01
Training a FFNN using Matlab Neural Networks Toolbox (NNT)
Y
ou were shown a pattern, where P is the input and T is the output. You were asked to remember it.
Since you only have about 10
11
to 10
12
neurons, you said, “Nah, I have better things to think about.
H
owever, I will program 6 neurons to do this.” So you configure a 2-layer 1-5-1 FFNN as follows:
( )
2
2
2 2 2 2
2 2 1 21 2 211 212 213 214 215
11
12
1 1 1 13
14
15
11
12
1 13
14
1
2
( )=purelin( )
+
tan ( )
tan ( )
( )= tan ( )
tan ( )
tan ( )
J T y
y f s s
s b w w w w w
sig s
sig s
sig s
sig s
sig s
s
s
s
s
= −
=
= =
=
=
W y W
y f s
s
111 1
1
121 12
1 1 1 131 1 13
141 14
15 151 15
+
w b
w b
u w u P b
w b
s w b
= = = =
W b W b
0 1 2 3 4 5 6 7 8 9 10
0
0.5
1
1.5
2
2.5
3
3.5
4
Input P
Output T
1
1
f
∑
1
b
1
s
u P=
1
W
1
2
f
∑
2
b
2
s
2
W
1
y
2
y
y
2
should
reproduce the
pattern output T
when input P is
shown at FFNN u,