How to remove for loops from a code

I've to remove for loops from a code to speed it up.

the code is:

function [S p] = similarityMatrix(data,candidates)

[row,col] = size(data);
S = zeros(row,col,row*col);
if nargin == 1
    candidates = 1:numel(data);
end
for z = 1 : numel(candidates)
    c = candidates(z);
    for i = 1 : row
        for j = 1 : col
            if c == sub2ind(size(data),i,j);
                S(i,j,z) = NaN;
            else
                S(i,j,z) = -dist(data(c),data(i,j));
            end
        end
    end
end
%aggiungo le preferenze
p = nanmedian(S(:));
S(isnan(S)) = p;

is there a way to improve it?

Answer 1

[Andrei Bobrov]

n = numel(data);
S = -abs(bsxfun(@minus,data,reshape(data(candidates),1,1,[])));
S(candidates + (0:numel(candidates)-1)*n) = nan;

p = nanmedian(S(:));
S(isnan(S)) = p;