How can i optimize the running time in for loops

I' trying to compute the Heterogeneous Value Difference Metric distance between objects in numerical data set, I'm using my own code

 for i = 1:size(num_data,1)
        count = 0;
        for j = 1:size(num_data,1)
            sim = 0;
            for k = 1:size(num_data,2)
                sim = sim + ((abs(num_data(i,k) - num_data(j,k)))/...
                    (max(num_data(:,k))- min(num_data(:,k))))^2;
            end
            sim = sqrt(sim);
            if sim <= threshold
                count = count + 1;
                numclass(i,count) = j;
            end        
        end
  end

where num_data is the numerical data set, but my problem is that the previous code takes too long to run especially when the data set contains a lot of objects. At the same time I'm unable to use parfor because i have dependencies in the loops so, any suggestions to improve the running time.