I just asked the following question:

What are attention mechanism exactly?

It seems that attention-related questions haven't received answers on this site. Is this the right site for these questions? After Data Science, I found another site which seems related to Deep Learning questions: Artificial Intelligence (and I hope there isn't a third one). I believe Cross Validated is the most suitable site for this kind of questions, but I would like to have your confirmation.

  • 3
    $\begingroup$ It seems to pertain to ANNs, which are on topic. I would say it's OK. We can see how it fares. In the interim, you added the [attention] tag, but it has no excerpt. Can you provide one? $\endgroup$ May 4 '18 at 19:03
  • 3
    $\begingroup$ Hi @gung, I didn't create the [attention] label, I just used it because the site auto-completion suggested it (I did notice that it didn't have an excerpt, which I found odd). Given that I've never really understood what attention mechanism are (that's why I'm asking), I don't feel qualified to write an excerpt. If no one else volunteers to write a proper excerpt, I can surely parrot back the sentences commonly used to describe attention, but I'm not sure I'll make the community a great service. $\endgroup$
    – DeltaIV
    May 5 '18 at 10:42
  • $\begingroup$ FWIW the tag was created by @FranckDernoncourt. $\endgroup$
    – amoeba
    May 5 '18 at 12:17
  • $\begingroup$ Sorry for not having written an excerpt when creating the tag :-) $\endgroup$ May 17 '18 at 4:42

It's on-topic here because it is about machine learning, and machine learning is squarely on-topic here.

  • $\begingroup$ Some folks have said "it is too software-side of machine learning" so it goes on stack-overflow. Some folks have said "its too pure-mathy" so it goes on mathematics or data science. Some have said "its too academia" or "its too work-professionalism" so I had to move to those forums. There is no perfect Venn diagram as to which parts of stack-exchange it goes on, but it can go on one of them. $\endgroup$ May 10 '18 at 18:28
  • 4
    $\begingroup$ @EngrStudent those folks are wrong. StackExchange does not have the philosophy "It's not offtopic here only because it's better suited elsewhere". If something is on-topic on both A and B, OP can decide where to post, but only pick one. $\endgroup$
    – FooBar
    May 15 '18 at 12:14
  • $\begingroup$ Try this: stats.stackexchange.com/review/close/186430 Since when isn't using a neural network to forecast time-series data on-topic? $\endgroup$ May 17 '18 at 14:19
  • 1
    $\begingroup$ @EngrStudent I'm not sure what you're asking. In this thread, DeltaIV asks about a specific question, and I say that DeltaIV's question is on-topic because it's about ML. You're asking about a different question, and saying that this second question is on-topic because it's about ANN time-series forecasting. This seems to be an entirely new question. Perhaps you would find more helpful answers if you opened your own meta thread. $\endgroup$
    – Sycorax Mod
    May 17 '18 at 14:32
  • $\begingroup$ I was just supplying one of several examples where just because something is "machine learning" or otherwise "core" to stats, it gets voted as "off-topic" for the stats forum. Goes along with the first topic. fwiw, I love your answer, and I want these sorts of things to be on-topic. I've just been axed and seen things axed that depart from the ideal. $\endgroup$ May 17 '18 at 14:55
  • 2
    $\begingroup$ @EngrStudent Just speculating, but when the title includes "how to do ... in R," people tend to regard that as off-topic because the question asks for explanation of how to use software -- that is, the answer is code and/or software documentation. On the other hand, if the question asked for the predictive equations for a specific model, it would probably be on-topic -- the answer is about machine learning as a mathematical discipline, not software engineering. $\endgroup$
    – Sycorax Mod
    May 17 '18 at 14:58

You must log in to answer this question.

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