You need to re-evaluate how you are doing things. In the first place, you have the Euclidean calculation wrong. It should be this using your formulation:
d = sqrt(sum((Point_Data(i1,:)-Point_Data(j1,:)).^2));
or maybe just this more simply:
d = norm(Point_Data(i1,:)-Point_Data(j1,:));
And your test should be d>4 instead of d<4.
Secondly, your code iteratively fills up a sparse matrix. At each iteration when you fill an element, the entire data set must be copied into newly allocated memory to make room for the new element. This results in a serious drag on performance. For a large sparse matrix it would not be unusual to see iterative size increasing code to take days or even weeks of computation time. There are better ways of computing the Euclidean distances between sets of points. If you do end up using for loops, the general advice is to save the i,j,v triples off to the side and then build the sparse matrix all at once from these triples.
But most importantly, the size of the result may overwhelm your memory. Before you even try to build such a result, you need to figure out if the result is even something you can handle. In your particular case, what is the expected sparsity of the result? I.e., have you calculated the memory requirements of your expected result, and can you handle it? And what do you intend to do with this result downstream in your code? Maybe there are better ways of getting your problem solved rather than forming this potentially very large matrix explicitly.
