2
$\begingroup$

I have a large symmetrical dissimilarity matrix of dimension 300 000. Can you please suggest the multidimensional scaling algorithms that can work with such large data? Input of course can be the original vectors from which the dissimilarity matrix was calculated.

$\endgroup$
2
  • $\begingroup$ 300000x300000 matrix? Wow. Can your computer keep it in RAM memory? Honestly, I doubt that anyboby in the world has ever done MDS on such data. You'll probably be the pioneer. $\endgroup$
    – ttnphns
    Nov 17, 2013 at 8:05
  • $\begingroup$ it's not a new task, there are a lot of publications on this, but very little accessible code. $\endgroup$ Nov 17, 2013 at 8:43

1 Answer 1

2
$\begingroup$

You can try MDS methods like Sammon, CCA, tSNE or a lot of other non-linear dimensionality reduction methods. The trick is that you don't calculate the dissimilarity matrix, but just calculate individual distance on-fly in the MDS methods.

$\endgroup$
1
  • $\begingroup$ James, if you can - please give links to descriptions of these methods. $\endgroup$
    – ttnphns
    Nov 18, 2013 at 8:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.