Matlab code about cosine similarity

n_row = size(A,1);
norm_r = sqrt(sum(abs(A).^2,2)); % same as norm(A,2,'rows')
B = zeros(n_row,n_row);
for i = 1:n_row
  for j = i:n_row
    B(i,j) = dot(A(i,:), A(j,:)) / (norm_r(i) * norm_r(j));
    B(j,i) = B(i,j);
  end
end

If you know The improved Cosine similarity , Welcome to exchange ~