6

The question: https://stackoverflow.com/questions/14939245/can-a-neural-network-provide-more-than-yes-or-no-answers

~1,100 other (unclosed) questions about neural networks: https://stackoverflow.com/search?q=neural+networks+closed%3Ano

It says it was closed as "Off Topic". Yet when I follow the FAQ that the closed message itself links, it says the following:

We feel the best Stack Overflow questions have a bit of source code in them, but if your question generally covers …

  • a specific programming problem
  • a software algorithm
  • software tools commonly used by programmers
  • practical, answerable problems that are unique to the programming profession

… then you’re in the right place to ask your question!

It also says:

You should only ask practical, answerable questions based on actual problems that you face.

To my understanding, the question closed meets all of these criteria. It is a question about a specific kind of software algorithm (multilayer perceptrons) and its limitations. It was also practical and answerable, as demonstrated by the fact that other people were facing the problem and it was answered.

Was this close justified? If so, can you explain why?

Addendum: Apparently it's not obvious from the question-verbiage based on comments below, so just to make it clear: I'm not asking for the question to be reopened. I would just like to understand the rationale.

Improve this question
20
  • 5
    Sounds better for the whiteboard of Programmers
    – random Mod
    Commented Feb 19, 2013 at 21:13
  • 4
    "~2,500 other (unclosed) questions about neural networks"..that does not necessarily mean anything with regards to your question.
    – Bart
    Commented Feb 19, 2013 at 21:14
  • Note that a question must meet the criteria you quoted from the FAQ to belong on SO, but not all question that meet those criteria belong on SO. There are a number of other criteria discussed throughout the rest of the FAQ.
    – Servy
    Commented Feb 19, 2013 at 21:16
  • @Bart The question was closed because it was about neural networks. Other neural networks questions abound on the website. Their existence and status is clearly relevant.
    – asteri
    Commented Feb 19, 2013 at 21:16
  • 4
    @Jeff "The question was closed because it was about neural networks." What's your basis for that assertion? If there are a lot of on topic neural network questions that may not be true, therefore it must have been closed for some other reason. I can tell you that you'll find the meta community much more receptive if you ask why the question was closed, rather than saying it shouldn't be and demanding it be reopened.
    – Servy
    Commented Feb 19, 2013 at 21:16
  • @Jeff Of those questions only 1095 are not closed. But still, not necessarily of relevance to your question.
    – Bart
    Commented Feb 19, 2013 at 21:17
  • 1
    @Servy The reason it was closed was listed as "Off Topic". The topic was neural networks.
    – asteri
    Commented Feb 19, 2013 at 21:17
  • 1
    @Servy I am not demanding it be reopened. I don't care if it's reopened; some fine gentleman already gave me the information I wanted. I'm asking what the rationale for it being closed was. Hence the questions at the end. It is a "Why" question, not a demand. I thought that was self-explanatory.
    – asteri
    Commented Feb 19, 2013 at 21:19
  • @Jeff That's one of the broader topics of the question. The question also discusses other more specific topics, or can be of a form that is out of scope for the site, etc. You are thinking too narrowly.
    – Servy
    Commented Feb 19, 2013 at 21:19
  • 1
    @Bart With respect, I feel as though you are being purposefully dense and antagonistic for no real reason. It's obvious that the status of other questions of the same topic are relevant to the discussion of an off-topic closure.
    – asteri
    Commented Feb 19, 2013 at 21:29
  • 2
    @Jeff A question about C# can be off topic on Stack Overflow, although ostensibly C# is a topic very relevant to Stack Overflow. Just because your question tagged neural-network was closed as off topic doesn't mean the community feels neural networks are as a whole off topic on SO.
    – user200500
    Commented Feb 19, 2013 at 21:30
  • 2
    @Jeff Mwah, that does not show evidence of a whole lot of respect. But disregarding that, when asking for an explanation of closure (or even when asking to reopen a question), the whole line of argument of "this large set of questions was not closed" does not necessarily mean a whole lot when those questions either should have been closed as well but weren't, or when their actual content (and not a mere keyword) is significantly different. Focus on your own question and its merits. Don't drag in a random number of other questions. That's all I'm saying.
    – Bart
    Commented Feb 19, 2013 at 21:31
  • 2
    That said, it is always unfortunate when a question is closed without even a single comment explaining the rationale, since it leaves both the OP and other users guessing about how the question was off topic.
    – user200500
    Commented Feb 19, 2013 at 21:38
  • @Servy, understood. However any sub-topic of the question is also on topic according to the FAQ linked as a close reason. Back propagation of error is a specific algorithm. Perceptions are a specific neural network implementation. Image recognition is a common programming problem. No matter which topic you extrapolate to be the "main" one, it seems on topic for SO. Hence the question.
    – asteri
    Commented Feb 19, 2013 at 21:59
  • Also above @Asad
    – asteri
    Commented Feb 19, 2013 at 22:00

1 Answer 1

Reset to default
5

The question itself was off topic for Stack Overflow, because it spoke in generalities, and not in an actual practical problem you face... With code.

It's a great question, it's just off topic for Stack Overflow. That's why it was closed. The people who closed it could have thought it was on-topic for Programmers -- but normal users can't migrate something to programmers.

I've migrated it to programmers since it's much more of a white-boardy question. If you have a specific implementation that's returning results you wouldn't expect, the that would be on topic for Stack Overflow. Just show us your research.

The problem isn't as much the language-agnostic tag is a meta tag and should be abolished as it is that your question lacked a specific problem you faced.

From the FAQ:

Stack Overflow is for professional and enthusiast programmers, people who write code because they love it. We feel the best Stack Overflow questions have a bit of source code in them, but if your question generally covers …

  • a specific programming problem

And then later:

You should only ask practical, answerable questions based on actual problems that you face.

The emphasis is on a 'practical' question based on an actual problem you face. Most of the time this is apparent because there exists code to help us solve that actual problem you face. When there isn't code, we determine (based on the content of the question) if it's a whiteboard-y question. If so, it goes to Programmers.SE. If not, it gets closed as off-topic.

Improve this answer
6
  • Thanks for your answer. So is it the case that language agnostic questions are "off-topic" and all questions should be related to specific code samples? If so, is this something that should be added to the FAQ?
    – asteri
    Commented Feb 20, 2013 at 13:48
  • @Jeff I updated my answer to address your comment. Short answer, No.
    – George Stocker Mod
    Commented Feb 20, 2013 at 14:51
  • @Jeff: your question was "off topic" because you had no real problem (nb: when I say "real" I mean software engineering real, programmers has a wishy-washy definition). I personally would have voted "Not Constructive" because a valid answer to your question is "Yes" (simple logic reveals this to be true even at the basic backprop ANN level: 2 output nodes represents 4 states...QED).
    – user7116
    Commented Feb 20, 2013 at 19:16
  • @sixlettervariables Thanks for your input. While I agree that superficially "simple logic" is all that's required to answer the question, the details provided in the question body showed a little more depth of thinking about whether backpropogation of error as an algorithm was capable of accurately producing non-boolean answers.
    – asteri
    Commented Feb 20, 2013 at 20:15
  • @Jeff: I don't follow. Backprop wouldn't work at all if it couldn't work with N output nodes (hidden nodes are just output nodes, QED). Granted, your problem space would have to be linearly separable, but that is another story (AND, OR, but not XOR with 1 hidden layer...so add another layer).
    – user7116
    Commented Feb 20, 2013 at 20:26
  • @sixlettervariables standoutpublishing.com/g/catastrophic-forgetting.html I didn't know this "problem" was already predefined, but it's what I had in mind when I asked the question.
    – asteri
    Commented Feb 20, 2013 at 23:00

You must log in to answer this question.

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