Euclidean Distance K-Means Clustering

distkmeans.m

function [U,V,idx] = distkmeans(X,k,tol,imax)

  [d,n] = size(X);
  
  U = zeros(d,k);
  V = zeros(k,n);
  
  % random clusters.
  for j = 1:n
    V(randi(k),j) = 1;
  end
  
  % monitor convergence.
  olderr = 0;
  
  for iter = 0:imax
    % recompute assignments.
    if iter > 0
      for i = 1:n
        dists   = sum(bsxfun(@minus, U, X(:,i)).^2, 1);
        [~, ix] = min(dists);
        col     = zeros(k,1);
        col(ix) = 1;
        V(:,i)  = col;
      end
    end
    
    % monitor cost.
    newerr = 0;
    
    % update or initialize cluster vectors.
    for i = 1:k
      ix      = find(V(i,:) > 0);
      c       = mean(X(:,ix), 2);
      U(:,i) 	= c;
      
      for ixi = 1:numel(ix)
        newerr = newerr + sum((X(:,ix(ixi)) - c).^2);
      end
    end

    newerr = sqrt(newerr);
    
    fprintf('Iteration %f\tCost function: %f\n', iter, newerr);
    
    if newerr > 0 && olderr > 0 && olderr - newerr < tol
      break;
    end
    
    olderr = newerr;
  end
  
  % which clusters?
  [~,idx] = max(V);
  
end