function [P,C,Err]=kmeansUpdate(X,C)
% function C=kmeansUpdate(X,C)
%  Performs 1 step of the k-means algorithm where X is p by n (data)
%  and C is k by n (k clusters)
% P holds the sorted data 
% C is the new matrix of clusters
% Err is a vector of distortion errors

%Check to see if dimensions are correct:
[m,n]=size(X);
[k,p]=size(C);
if n~=p
    error('Data dimensions do not match in kmeansUpdate\n');
end

% Compute the distances from all the points to each center
D=edm(X,C);

% Find the minimum distance for each and sort (Idx is the vector M from the
% example in the text).
[Vals,Idx]=min(D,[],2);
Err=zeros(k,1);         %To hold the distortion errors.

% Resort the data to get the new centers.
for j=1:k
    idx=find(Idx==j);
    if length(idx)>0
        P{j}=X(idx,:);
        C(j,:)=mean(P{j});
        Err(j)=Err(j)+sum(Vals(idx));
    end
end