Introduction to Pattern Recognition: A Matlab Approach

2.8 Multilayer Perceptrons 71

(a)

20

(b)

210 25 0 5 10 15 20

210

25

0

5

10

15

210 25 0 5 10 15 20

28

26

24

22

0

2

4

6

8

Class 11 training vector

Class 21 training vector

FIGURE 2.13

(a) Training set for Example 2.8.1. (b) Decision regions formed by the two-layer FNN with four hidden-layer

nodes, trained with the standard BP algorithm with learning rate 0.01 and 9000 iterations.

% Generation of training data from the second class

N2=80; %Number of second class data points

for i=1:c2

S2(:,:,i)=s*eye(l);

end

sed=0; %Random generator seed

[X2,y2]=mixt_model(m2,S2,P2,N2,sed);

%%%%%%%%%%%%%%%%%

%Production of the unified data set

X1=[X1 X2]; %Data vectors

y1=[ones(1,N1) -ones(1,N2 ...