Daniel@0: function [S, mu, sigma2] = standardize(M, mu, sigma2)
Daniel@0: % function S = standardize(M, mu, sigma2)
Daniel@0: % Make each column of M be zero mean, std 1.
Daniel@0: % Thus each row is scaled separately.
Daniel@0: %
Daniel@0: % If mu, sigma2 are omitted, they are computed from M
Daniel@0: 
Daniel@0: M = double(M);
Daniel@0: if nargin < 2
Daniel@0:   mu = mean(M,2);
Daniel@0:   sigma2 = std(M,0,2);
Daniel@0:   sigma2 = sigma2 + eps*(sigma2==0);
Daniel@0: end
Daniel@0: 
Daniel@0: [nrows ncols] = size(M);
Daniel@0: S = M - repmat(mu(:), [1 ncols]);
Daniel@0: S = S ./ repmat(sigma2, [1 ncols]);
Daniel@0: