calculate centroid after kmeans

Hi,

Is it possible to retrospecively compute the centroids from the membership indices and the data after running kmeans?

Thanks!

Hans

Answer 1

[Image Analyst]

Why not simply use the centroid locations that kmeans returns to you? Did you know that kmean() returns those if you ask for (accept) two return values

[indexes, centroids] = kmeans(data, k);

[Hans van der Horn]

Thanks for your answer.

I'm aware of the answer, but was looking for a way to retrieve the centroid in case I didn't ask for that output.

Any ideas?

[Hans van der Horn]

I already figured it out. In kmeans code for correlation (distance method I use) it's the following code for each cluster:

X = bsxfun(@minus, X, mean(X,2));

Xnorm = sqrt(sum(X.^2, 2));

if any(min(Xnorm) <= eps(max(Xnorm)))

error(message('stats:kmeans:ConstantDataForCorr'));

end

X = bsxfun(@rdivide,X,Xnorm);