function [alpha,b,w]=smo_svm()
MAX = 100;
ep = 1e-2;
thre = 1e-3;
data = dlmread('SMO_data.txt');
L = size(data);
N = L(1);
n = find(data(:,L(2))==0);
data(n,L(2)) = -1;
y = data(1:N,L(2));
x = data(1:N,1:L(2)-1);
y = data(1:N,L(2));
% alpha = rand(N-1,1);
% alpha(N) = -y(N)*sum(alpha.*y(N-1));
C = 10;
alpha = [0; C; C; zeros(N-3,1)];
b = 0;I1_old = 0;
test = 0;
iter = 0;
while (test==0 && iter<MAX)
iter = iter+1;fprintf('Iteration Number: %f\n', iter);
w = CreateWeights(x,y,alpha);
KKT = CalculateKKT(x,y,alpha,b,w);
temp = find(KKT == max(KKT));
I1 = temp(1);
if (I1 == I1_old)
I1 = ceil(N*rand);
end
I1_old = I1;
x1 = x(I1,:);
for i = 1:N
e(i) = CalculateE(I1(1),x1,x,b,y,alpha)- CalculateE(i,x(i,:),x,b,y,alpha);
end
temp = find(abs(e) == max(abs(e)));
I2 = temp(1);
x2 = x(I2(1),:);
k = sum(x(I1(1),:).*x(I1(1),:))+sum(x(I2(1),:).*x(I2(1),:))-2*sum(x(I1(1),:).*x(I2(1),:));
if (k <= 0)
continue;
end
temp = alpha(I2(1));
if(y(I1)~=y(I2))
U = max(0,alpha(I2)-alpha(I1));
V = min(C,C+alpha(I2)-alpha(I1));
else
U = max(C,-C+alpha(I2)+alpha(I1));
V = min(0,alpha(I2)+alpha(I1));
end
if (U == V)
continue;
end
alpha(I2(1)) = alpha(I2(1))+y(I2(1))*CalculateE(I2(1),x2,x,b,y,alpha)/k;
if (alpha(I2)>V

1. The figure below depicts the linear discriminant g(x) as discuss...