Scribd
Upload
38 views

Ee133A Hw1-Matlab

The document describes performing k-means clustering on handwritten digit data. It initializes cluster centers randomly, assigns digits to the closest center, recomputes the centers as the mean of each cluster, and iterates this process 100 times. It prints the value of the cost function J at each iteration, with the final value of J = 4.013349e+01 in bold. Finally, it displays the images of the five learned cluster centers.

Uploaded by

Dylan Ler
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
38 views

Ee133A Hw1-Matlab

The document describes performing k-means clustering on handwritten digit data. It initializes cluster centers randomly, assigns digits to the closest center, recomputes the centers as the mean of each cluster, and iterates this process 100 times. It prints the value of the cost function J at each iteration, with the final value of J = 4.013349e+01 in bold. Finally, it displays the images of the five learned cluster centers.

Uploaded by

Dylan Ler
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

EE133A HW1-MATLAB

I = find(labels < 5);


digits = digits(:,I);
labels = labels(:,I);%only condier digits 0-4,reduce # of columns
[n,N]=size(digits);%28*28 image for each col of digits--size
J0=NaN;
z = rand(n,5);%generate initial values
Dis=zeros(5,N);
for iter=1:100
for k=1:5
Dis(k,:)=sqrt(sum((digits-z(:,k*ones(1,N))).^2));
end;
[d,classnew]=min(Dis);
if iter >1
if J1==J0,
break;
end;
J0 = J1;
end;
J1=norm(d)^2*(1/N);
disp(sprintf('J= %e.',J1));
for p = 1:5
I=find(classnew==p);
if size(I,2)
z(:,p)=sum(digits(:,I),2)/size(I,2);
else disp(sprintf('cluster %d is empty.',p));
return;
end
end

end
for k = 1:5
figure()
x=reshape(z(:,k),28,28);
imshow(1-x);
end

The following data was collected from iterations and the final value of J is in
bold, which is the smallest one as we desired.
J= 2.431542e+02.
J= 4.675040e+01.
J= 4.280635e+01.
J= 4.161833e+01.
J= 4.085309e+01.
J= 4.042314e+01.
J= 4.017846e+01.

J= 4.014629e+01.
J= 4.013801e+01.
J= 4.013514e+01.
J= 4.013412e+01.
J= 4.013369e+01.
J= 4.013355e+01.
J= 4.013352e+01.
J= 4.013350e+01.
J= 4.013349e+01.
J= 4.013349e+01.
J= 4.013349e+01.
J= 4.013349e+01.
J= 4.013349e+01.
The following graphs show the five representatives:

You might also like