I've seen at least 2 questions that were closed as too broad but were specific and answerable within their subdomain. And in both cases, they've failed to reopen appropriately.

One of the examples is linked below which I'll reference. The original question was a little hard to follow, and as far as I can tell the close votes just didn't understand it. It was my field of research though and I was able to easily answer it, and in fact, it was a critical technical question in a rather specific subfield of machine learning.

I answered the question in comments, requested that it get reopened, and the OP changed the question to be more clear to a broader audience.

It's received 4 reopen votes and has been hung there for many days. It seems like 5 reopen votes is too many and I propose that a more reasonable number of reopen votes is 3.

Multi-Task Learning: Train a neural network to have different loss functions for the two classes?

  • 9
    The question doesn't have a single line of code. The question also is seeking the best way, which is an extremely broad statement, and will result in opionionated answers in my opinion. – Ramhound Jan 24 '18 at 15:02
  • 3
    the edit of that question made it worse, not better. – rene Jan 24 '18 at 15:04
  • 4
    So you think that the question is unclear, and yet are somehow upset that it's closed. We want unclear questions to be closed. That's why there's a close reason just for it. If people didn't understand what it was asking then that's a problem with the question that needs to be fixed before it can be answered. – Servy Jan 24 '18 at 15:04
  • I'm afraid your example is drawing "belong to specific site" close votes, better get rid of it. Instead, focus on why you think this feature request will be useful. – Shadow The Vaccinated Wizard Jan 24 '18 at 15:07
  • Is it unreasonable or not practical to have code in that question? Or are the things like nan, loss, regressed feature obvious terms based on the textual description alone? – rene Jan 24 '18 at 15:10
  • 1
    It didn't survive the re-open queue: stackoverflow.com/review/reopen/18592789 – rene Jan 24 '18 at 15:11
  • I hesitate to say I'm upset, I'd rather not describe it in terms of emotions. I found this question answerable (I'm providing the answer), and I felt it was an important and useful question. And I answered it in comments. This was not the first question that this occurred on. In a separate case there was a specific question about a specific function in tensorflow that was closed in much the same manner. The fact that I answered it in comments and asked for it to be reopened there wasn't sufficient to reopen it (I can't find that link just now), and I just found that to be a bad process. – davidparks21 Jan 24 '18 at 17:27
  • 2
    If anything, asking in a comment to have a question re-opened is useless. Make an edit if you can that clarifies and/or improves the question so the re-open reviewers have something to judge. Being able to answer a question doesn't make the question on-topic / useful etc. On SO you're free to hop in the SOCVR chatroom if you need advice or help to get questions closed or re-opened but we still need some argumentation why a post needs moderation. If specific type of questions need special care, post on the site meta to get consensus and guidance for such questions. – rene Jan 24 '18 at 18:23

I don't think this is a good idea to reduce the amount of reopen votes required to be only 3.

Reopening should be the same level as closing.

If it requires 5 users to close, it should require 5 users to reopen.

That said, I would support extending Increase close vote weight for gold tag badge holders to be more than just for duplicates, for example an off topic close vote of a user with gold tag badge will be "worth" two ordinary close votes, so it will take 3 such users to close/reopen questions. (or actually two, plus one "ordinary" high rep user.)

  • This would be a reasonable approach to me. Or perhaps provide an option to still answer the question. I do question a process that fails to reopen the question when an answer is clearly made available. – davidparks21 Jan 24 '18 at 17:32
  • @davidparks21 we can still post answers to closed questions about 2 hours after they're closed. Other than that, if a question is clearly on topic, you can try get some help in chat, there are many high rep users roaming various rooms. – Shadow The Vaccinated Wizard Jan 24 '18 at 20:42
  • Note that I'm the one answering the question, not the OP. Sounds like that might not have been clear. – davidparks21 Jan 24 '18 at 23:44
  • @davidparks21 this is not relevant, my answer is for the general case. – Shadow The Vaccinated Wizard Jan 25 '18 at 0:04

I voted to close the question because, the way I see it, the lack of actual content accompanying the description of the context and the intended goal makes it quite open-ended. We know that it is a question about Keras, and that the OP built a model with two loss functions (I'll be ignoring the previous revisions, since those had even less detail):

I have a neural net with two loss functions, one is binary cross entropy for the 2 classes, and another is a regression. Now I want the regression loss to be evaluated only for class_2, and return 0 for class_1, because the regressed feature is meaningless for class_1.

Yet, we were given no MCVE to work with. It would have been much better to have some code depicting how the neural network is currently created and trained, even if the code had the issue mentioned afterwards:

Training it separately on only class_1 data doesn't work because I get nan loss.

Without the MCVE, it is not something that we can easily reproduce (no data, no model specifications, no optimizer algorithm, no hyperparameters, ...). The answerer would have to fill in the gaps. And so, the question in its current state looks like a code request for a generic solution to multi-task learning in Keras.

At the time of writing this, there is one reopen vote left to lift the closure. In my opinion, the question should be improved with a proper MCVE before it's reopened. Changing the number of reopen votes necessary not only feels unfair towards close votes, but would also not be healthy to question subdomains with less moderation. In particular, I found the situation in the context of deep learning questions to be alarming enough to raise a meta question many months ago. This is not to be interpreted as some form of spite towards deep learning questions, but a suggestion to make them better (or make them elsewhere if they're off-topic).

  • I'm not trying to blast the voters by any means, the first version of it was quite hard to follow, and had it not been an area of specific interest to me I would have never caught the intent of the question. Your point that it might better reside somewhere like DataScience stack exchange may be quite pertinent, tensorflow questions tend to come to SO, but tensorflow overlaps data science quite a lot. Though I still question the reopen process when an answer is clearly made available. – davidparks21 Jan 24 '18 at 17:30
  • @davidparks21 I understand that answers may become available to these questions. However, that is a double-edged sword... in which the edge closest to the wielder is sharper. Analogies aside, they tend to attract low quality answers, and so we close them to amend that. Whether it should be opened or not, that should depend on the question, and not just one potential answer. And please note, my final statement in parentheses is not applicable here: I find the question to be sufficiently on-topic for Stack Overflow, but it needs significant improvements to become salvageable. – E_net4 the flagger Jan 24 '18 at 18:01

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