From meta.SE, Guidance for asking "Tune my machine learning model" questions asks whether such questions are on-topic at any network sites, and naturally ours is one of the possibilities. So, (when) are they on-topic here?
$\begingroup$
$\endgroup$
2
-
2$\begingroup$ I tend to agree with Scortchi's comment and answer there, that aside from generalizing them a lot such questions are unlikely to be useful beyond the asker. But sometimes there can be a common gotcha with a particular model or problem archetype, and in those cases I think it could make a good Q&A; I'm not sure how to draw that line though. $\endgroup$Ben Reiniger– Ben Reiniger Mod2024-08-22 00:43:08 +00:00Commented Aug 22, 2024 at 0:43
-
$\begingroup$ There are currently 44 questions containing "improve performance", but most of them aren't this sort of question I think, and of course that search is missing some questions of this sort. Most of them aren't upvoted and/or don't have answers anyway. Maybe the closest to a good example is datascience.stackexchange.com/q/100674/55122, datascience.stackexchange.com/q/74118/55122? $\endgroup$Ben Reiniger– Ben Reiniger Mod2024-08-27 13:47:05 +00:00Commented Aug 27, 2024 at 13:47
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
I'll propose that these are on-topic, but caution askers that such questions are often a poor fit to the Stack Exchange network, and may not attract many answers.
- They're too broad. They can work toward the "repository of high-quality answers" mission of the network if they collect multiple answers that may help future readers with various issues that look like yours, but notably those may not help you the asker with your problem.
- Minimal reproducible examples, while not strictly required here as on Stack Overflow, are very helpful to potential answerers, but very hard to provide in full for these sorts of problems (taking a sample is likely not to show the full issue).
- The answers may be subjective, especially if the goal is some vague performance boost.