I don't really understand if SVM are classification methods (like Logistic regression) or clustering methods.

Since it's used for supervised learning, it should be part of classification methods right ?

Or can SVM also be used for unsupervised learning, and would then also be part of clusering methods ?


SVM are one of the most widely known classifiers.

There also exists SVR, Support Vector Regression.

As SVMs require training and hyperparaneter optimization they are only suited for supervised learning, and cannot be used for hard problems such as clustering.

  • $\begingroup$ What about support vector clustering? $\endgroup$ – SmallChess Mar 17 '17 at 9:16
  • $\begingroup$ I'm not aware of anybody using it on real data. It's very expensive (so it won't work on larger data) and you need excessive number of runs, need to magically choose hyperparameters etc. - I fear that the reported results are essentially due to overfitting. $\endgroup$ – Anony-Mousse Mar 18 '17 at 19:49

SVM are one of the most widely used "classifiers". however you can also do regression tasks but that is called support vector regression.it is better to call them "kernel based methods". there is a very little work done in unsupervised domain in svms (one class svms).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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