10
$\begingroup$

When a question is put on hold as unclear (for example), is there a way to propose that it is reopened? Like flagging? In particular, I think this question is clear enough to be answered and I would like to post an answer:

Do Gaussian process (regression) have the universal approximation property?

$\endgroup$
  • 4
    $\begingroup$ There is a procedure but you need 3000 rep to participate $\endgroup$ – mdewey Mar 19 '17 at 16:17
  • 4
    $\begingroup$ Even when you're sure you know what's being asked, it's usually not a good idea to attempt to answer an ambiguous, nonsensical, or self-contradictory question. That can be worse than no answer at all, because other readers (including the original proposer) might interpret it in a different sense than you and completely misunderstand your answer. The comment thread to that question continues to suggest additional clarification is needed. When (if) the OP elects to edit the post, it will become available for consideration in the reopen queue. $\endgroup$ – whuber Mar 19 '17 at 18:12
  • $\begingroup$ @whuber, I can't say I like the idea of disagreeing with an user of your experience and skill, but this time I feel I have to. I mean, surely I agree that questions on hold as unclear shouldn't (and actually can't) be answered. But I don't think the question in object is that unclear. Universal approximation property is a term quite often used with Gaussian Processes, and in general for RKHS methods, both in a frequentist or in a Bayesian framework, although with admittedly quite different meanings. $\endgroup$ – DeltaIV Mar 19 '17 at 19:47
  • $\begingroup$ maybe I should delete the preceding comment - I don't want to open a technical discussion on Meta, which is not the right place for it. $\endgroup$ – DeltaIV Mar 19 '17 at 19:49
  • $\begingroup$ I'm not an expert in this, and would gladly defer to you about the subject, but I think I can recognize when a question is inconsistent or self-contradictory. In this case the referenced Wikipedia article is about neural nets and not about Gaussian processes or RKHS methods. Moreover, I am familiar with the standard definitions of "function" and "process": because the question appears to use these in an almost nonsensical way, I am comfortable with waiting until the OP can explain what he means. $\endgroup$ – whuber Mar 19 '17 at 20:06
  • $\begingroup$ I didn't want to offend anyone. I thought that my comment to the original question, containing a couple links to universal approximation theorems for some particular GP models, would help clarify the situation, but of course that's a personal opinion and I can be wrong. Apologies if I came out as offensive or exerting pressure - it was not intended. $\endgroup$ – DeltaIV Mar 19 '17 at 20:20
  • 8
    $\begingroup$ @DeltaIV I don't think there's a problem. You weren't rude or aggressive. It's fine to ask for clarification or explanation, and it's okay to politely disagree and explain why you think as you do. $\endgroup$ – Glen_b Mar 20 '17 at 5:14
  • 2
    $\begingroup$ DeltaIV, I agree with @Glen_b and apologize if (in the process of conforming to the severe word count limits of a comment) I left the impression of being offended or annoyed: that was not my intent. The question remains ambiguous and self-contradictory, and the OP has made no efforts so far to clarify it, so I will suggest caution: if you (or anyone else) might elect to answer it, please take care to state what you think the question actually asks before you offer an answer. $\endgroup$ – whuber Mar 20 '17 at 17:37
  • $\begingroup$ @whuber and Glen_b, thanks for the clarification. If I decide to answer the question, I will take care to clarify my interpretation of the question. $\endgroup$ – DeltaIV Mar 20 '17 at 17:50
2
$\begingroup$

I can think of in 4 options, which #1 and #2 are the best ones. They are:

  1. If the question was not already edited after it was put 'on hold'; edit it for clarification (if you are positive you know how), so it can be automatically sent to the reopen queue for peer review. However, a caveat is that in many occasions only OP will be able to edit the question to address the close reason (don't edit in this case; wait for OP to step in).

  2. Reach 3k reputation and earn cast close/reopen vote privilege, so to cast a reopen vote and send the question to the reopen queue for peer review. Leaving a comment explaining why you think the question should be reopened, might increase chances of it being reopened.

  3. Flag it for moderator attention and explain your reasons, tell you want to post an answer; but unless it is very good reason it won't work.

  4. Ask for help in CV Meta (kind of what you did with this post).

In my opinion #4 competes with #3. However, #1 or #2 should always be tried first, and in case of negative results, see if any new piece of information was added (e.g.: edits to question or new comments from peers), then evaluate if #3 or #4 are worthy a shot.

$\endgroup$

You must log in to answer this question.

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