3
$\begingroup$

This question already has an answer here:

I was reviewing the "kernel" tag and this tag may refer to several different items.

In statistics, Kernel has a particular meaning.

However, a commonly encountered use of the term kernel in the forum is related to the Kernel Trick, especially common in questions related to SVMs and Machine learning.

Is there a way to separate these meanings for the same tag? Or should a different tag "Kernel-Trick" be created?

$\endgroup$

marked as duplicate by amoeba, Dougal, gung, Nick Stauner, Scortchi Oct 19 '15 at 12:49

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

4
$\begingroup$

First of all, this raises a question how deep we should dive in decomposing this word -- even if you strip ML use, it is still pretty ambiguous.

The second issue is that it seems hard to maintain. New users usually heavily pollute enforced tag policies, and it is certainly not obvious that stamping SVM kernel question with [kernel] is wrong in any way.

$\endgroup$
  • $\begingroup$ Regarding the second issue, I agree: I see plenty of SVM kernel questions marked with [kernel] and this problem is definitely not obvious. Creating a new tag will have the downside of added moderator overhead (imagine fixing all those wrong [kernel] tags and replacing them with [kernel-trick] or [svm-kernel] tags...) $\endgroup$ – Andrew Apr 18 '12 at 7:57
  • $\begingroup$ I'm going to go ahead and create a new tag [kernel-trick] and we will see how it goes. With respect to non ML use, we might perhaps look into how to decompose that at a later stage. $\endgroup$ – Andrew Apr 21 '12 at 20:03
  • $\begingroup$ @Andrew, because svm + kernel imply kernel-trick, creating a new tag seems unnecessary. $\endgroup$ – whuber Apr 22 '12 at 21:03
  • $\begingroup$ Ok @whuber. Could you post that as an answer so that I accept and end this discussion? Unless anybody else adds something new or prefers another option it seems as if we have reached a satisfactory conclusion. Thanks. $\endgroup$ – Andrew Apr 22 '12 at 21:38
  • $\begingroup$ @Andrew, that comment is merely a clarification of what mbq has already said here, so accepting his reply would be fine. $\endgroup$ – whuber Apr 22 '12 at 21:41

Not the answer you're looking for? Browse other questions tagged .