Well, it took a while, but I'm finally done with this effort: every question in kernel has been sorted into kernel-trick and/or kernel-density-estimate.
kernel is now empty; a moderator should destroy it.
Synonymizing kernel-density-estimate to kernel-smoothing would remove the slight misclassification of lumping kernel regression in the KDE tag. I didn't count as I read them, but I would very roughly estimate that maybe 15% of questions currently in kernel-density-estimate are actually about kernel regression, and another reasonable portion are about kernel functions in general (not specific to KDE or regression). There are also a handful of questions in kernel-regression, which should also be synonymized to kernel-smoothing if that's what's happening.
The relevant tag wikis need to be updated. Since I don't have direct editing privileges on them, I figured I should wait to propose anything until exactly what's happening with the various kernel smoothing tags is finalized.
Edited to bump. It seems like no one now disagrees with the proposal of making both kernel-density-estimate and kernel-regression into synonyms of kernel-smoothing, and leaving kernel-trick as is. If anyone does have comments about that, please say so.
kernel-smoothing
+regression
would be much better. $\endgroup$kernel-smoothing
, since there's no density estimation going on in kernel regression, though I may not have said that clearly above.... $\endgroup$