16
$\begingroup$

This question was put on hold for being off topic with the following note:

"This question appears to be off-topic because EITHER it is not about statistics, machine learning, data analysis, data mining, or data visualization, OR it focuses on programming, debugging, or performing routine operations within a statistical computing platform."

First, my post was about neural networks in general, I just posted the Tensorflow code to illustrate my point, in the same way that multiple posts on this SE include the R code to illustrate a point, so I don't agree that the question "focuses on programming" solely.

Second if questions that dive into the details of the code are off topic, why do we have such tags as Tensorflow, R, Python, Scipy, etc...in the first place? Wouldn't any question with those tags be disqualified using the same criteria that my question was disqualified on?

Also can someone please migrate this question over to SO if it is indeed off topic?

$\endgroup$
  • 4
    $\begingroup$ A brief discussion among several of the mods indicated that we were all inclined to reopen the question; a sufficient consensus. $\endgroup$ – Glen_b Jun 11 '18 at 15:35
  • $\begingroup$ It's not a statistics question, it's strictly related to Tensorflow (and to an old version too, though it works perfectly in Tensorflow 1.9, because now we have 1000 different ways to do the same thing in Tensorflow: a brilliant software design choice, really). As a matter of fact, you couldn't have asked the same question in Keras or in PyTorch $\endgroup$ – DeltaIV Jun 12 '18 at 19:41
  • 3
    $\begingroup$ This frustrates me (too?): Separating statistics from software is about as useful as separating medicine from the medical instruments. You'll find SO has an equally frustrating door-in-the-face policy on questions. Here is a question about closing vs. migrating programming Qs. On CV, I close many such questions b/c they don't meet SO's reqs for a MRW example. But even SO preferentially passes Qs that are "interesting" to them. $\endgroup$ – AdamO Jun 12 '18 at 20:40
  • 3
    $\begingroup$ @AdamO I appreciate your frustration and understand where you are coming from. But is this particular question a good test case for litigating the case that we should accept more questions which are software-heavy? $\endgroup$ – Sycorax Jun 13 '18 at 14:35
  • 2
    $\begingroup$ Totally! I had similar tags on my questions. Discussing coding details should be valid at CV. Coding is the practical part of machine learning, besides the shiny theory. I can perfectly understand the theory and still make stupid (or less elegant) decisions in my implementation. This kind of best practices-recipes would increase the value of CV, IMHO. Maybe with a special tag? $\endgroup$ – Sören Jun 14 '18 at 15:52
  • $\begingroup$ +1. The fact that there are coding tags and mandating that questions be agonistic creates conflicting messages. $\endgroup$ – kilojoules Jun 19 '18 at 19:08
15
$\begingroup$

The actual criterion is in our help/on-topic (search for programming). A question containing code or asking about code -- or even sometimes asking for code is not automatically off topic, but many are. Briefly, if the issue is one that takes our subject-matter knowledge to answer it belongs here. If it doesn't (such as "what's wrong with this code?", which is nearly always about language issues*), then it's off topic.

[I feel that sometimes a few of our voters can be a little overenthusiastic in closing questions that contain code, but the form of the responses in comments combined with the fact that several of the close voters are machine learning experts suggest that the issue may be primarily one of coding. In any case, I'd hesitate to try to unilaterally overturn a decision by 5 of our high reputation users unless there was a clear indication that the question should stay.]

If you feel your post should be seen as on topic under our guidelines, the best thing to do would be to edit it to emphasize those aspects of the question and de-emphasize the code-related aspects.

If you want your post migrated, the right thing to do is flag it for migration. I'll leave it for now in case we get some consensus here that the question should be on topic after all, but you can choose to flag your post at any time; as long as it meets the criteria for being on topic at the destination and would not be likely to close as too broad, unclear, etc, then it will be migrated.

--

* and in any case, in that form rarely a good question for our site even when it is on topic

$\endgroup$
9
$\begingroup$

At the time that I voted to close the question, I interpreted it as asking "Why does this code do this thing?" which is a programming question and hence off-topic.*

After conferring with some of the moderators, I came around to the understanding that the core of the question is really whether or not an "epoch" is a concept that is relevant to obtaining predictions from a neural network that has already been trained, and the code snippet is more-or-less a curiosity.


* I've done a bit of work with TensorFlow but am by no means an expert and haven't ever used this function. I think that understanding why that code is written in that particular way and to what purpose would require asking the authors or a programming forum.

$\endgroup$

You must log in to answer this question.

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