1

I would like to use SVM of scikit-learn library to do unserpervised clustering. I have been reading the documentation and many links in the net, but I can't find how to do that. Would you mind explaining me how to use scikit-learn for that and also the concept of SVM unserpervised clustering?

southernKid33
  • 351
  • 1
  • 2
  • 14
  • While there was a publication on using SVMs for clustering, it never made it into practice. My guess is that it suffers from the need to do hyperparameter optimization, which you can't in unsupervised scenarios. Or mahbe it just didn't work, there are a lotmof publications that don't really work. – Has QUIT--Anony-Mousse Oct 06 '19 at 06:00
  • But my gut feeling is that it works at least with a lot of parameter fiddling and on synthetic data. But also that it is very slow. – Has QUIT--Anony-Mousse Oct 06 '19 at 06:42
  • Thank you. I found this, but there is not much info about that https://github.com/grantbaker/support-vector-clustering – southernKid33 Oct 07 '19 at 14:38
  • 1
    See the discussions there about hyperparameter tuning and sensitivity. The idea of this is interesting, but its practical applicability seems to be a big problem, so it never got used. – Has QUIT--Anony-Mousse Oct 07 '19 at 22:40

0 Answers0