I have a dataset where each sample has 187 dimensions. The normal class has 70.000 samples while the other 4 have at most 5.000. Ideally, I would like to keep at most 15000 samples from the normal class which are representative enough. In other words I want to discard redundant samples from the normal class keeping the most beneficial ones.
Has anyone tried something similar?

  • 1
    $\begingroup$ Hi, I would like to recommend you to use this code for your purpose: github.com/BeenKim/MMD-critic In this paper, they selected the representatives from a large amount of data. $\endgroup$ May 6, 2020 at 2:27
  • 1
    $\begingroup$ Thank you the paper seems useful and with a lot of relevant references included. $\endgroup$
    – Dimimal13
    May 7, 2020 at 20:02
  • $\begingroup$ Im glad it helps! $\endgroup$ May 8, 2020 at 4:55


Your Answer

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

Browse other questions tagged or ask your own question.