How to calculate distance from cluster centroids from the outputs of Matlab kmean() function

Question

I have 2 output clusters from k means matlab function
[idx,C] = kmeans(X,2);

I don't know how to calculate the distance between centroid and each point in cluster by using "idx"

I want to get matrix with all points that their distance to centroid >2

% not Matlab code; just illustrating concept    

example c1->{x1,x2}= x1-c1=3 x2-c1=2

c2->{y1,y2}=
y1-c2=4
y2-c2=1

output={y1,x1}

Answer 1

Try it this way:

Update The answer now uses loops.

r = randn(300,2)*5;
r(151:end,:) = r(151:end,:) + 15;

n_clusters = 2;

[idx, C] = kmeans(r, n_clusters);

clusters = cell(n_clusters, 1);
distances = cell(n_clusters, 1);
for ii = 1:n_clusters
    clusters{ii} = r(idx==ii, :);
    distances{ii} = sqrt(sum((clusters{ii}-C(ii,:)).^2,2));    
end

figure;
subplot(1,2,1);   
for ii = 1:n_clusters
    plot(clusters{ii}(:,1), clusters{ii}(:,2), '.');
    hold on
    plot(C(ii,1), C(ii,2), 'ko','MarkerFaceColor', 'w');
end

title('Clusters and centroids');

subplot(1,2,2);

for ii = 1:n_clusters
    plot(clusters{ii}(distances{ii} > 2,1), clusters{ii}(distances{ii} > 2,2), '.');
    hold on
    plot(C(ii,1), C(ii,2), 'ko','MarkerFaceColor', 'w');
end
title('Centroids and points with distance > 2');

To get a a cell with matrices with the points larger than 2, you can do:

distant_points = cell(n_clusters,1);
for ii = 1:n_clusters
    distant_points{ii} = clusters{ii}(distances{ii} > 2,:)
end