Classifier(Matlab).m

close('all');
clear;

m(:, 1) = [0 0]';
m(:, 2) = [7 7]';
S1 = 2 * eye(2);
S2 = 0.2 * eye(2);
P = [1/2 1/2];

% Generate X1 and the required class labels
N1 = 200;
randn('seed',0)

X1 = [mvnrnd(m(:,1), S1, fix(N1/2)); mvnrnd(m(:,2), S2, N1-fix(N1/2))]';
y1 = [(-1) * ones(1, fix(N1/2)) ones(1, N1-fix(N1/2))];

% Generate X2 and the required class labels
N2 = 200;
randn('seed',100)

X2=[mvnrnd(m(:,1), S1, fix(N2/2)); mvnrnd(m(:,2), S2, N2-fix(N2/2))]';
y2=[(-1) * ones(1, fix(N2/2)) ones(1, N2-fix(N2/2))];

% Compute the Bayesian classification error based on X2
S_true(:,:,1) = S1;
S_true(:,:,2) = S2;
[z]=bayes_classifier(m, S_true, P, X2);
bayY = 2 * z - 3;
err_Bayes_true = sum(bayY~=y2) / sum(N2)

% 2. Augment the data vectors of X1 
X1=[X1; ones(1,sum(N1))];

% Augment the data vectors of X2
X2=[X2; ones(1,sum(N2))];

% Compute the classification error of the LS classifier based on X2
[w]=SSErr(X1, y1, 0)
SSE_out=2*(w'*X2>0)-1;
err_SSE=sum(SSE_out.*y2<0) / sum(N2)

wX = -5:0.1:12;
wY = (-w(1) * wX - w(3)) / w(2);

figure(1), 
plot(X1(1, y1==-1), X1(2, y1==-1), 'bo', ...
     X1(1, y1==1), X1(2, y1==1), 'r.', ...
     wX, wY, 'g-');
figure(1), axis equal