If you look at the online courses, books, blogs on machine learning, many of them seem to focus on "programming the algorithms", often in those resources you see more Python code then mathematical formulas. Scholars like Andrew Ng seem to be telling us that anyone can be using machine learning and minimal math is required to do so. Does this mean we should adapt and be more open to programming-oriented machine learning questions ("help me fix my TensorFlow code" etc.)?

  • $\begingroup$ I am not sure how this is different from a thread from two days ago: stats.meta.stackexchange.com/questions/4990. $\endgroup$
    – amoeba
    Commented Oct 23, 2017 at 14:05
  • 1
    $\begingroup$ @amoeba this is more meta question if we should change our policies, rather then asking what are the policies. I agree with the linked thread, but maybe we should be even more open for the programming-like question. $\endgroup$
    – Tim
    Commented Oct 23, 2017 at 14:07
  • 8
    $\begingroup$ I don't see that we should. If various people are teaching machine learning via the pedagogical strategy of 'here's some code, don't worry about how or why it works', that seems like all the more reason we should focus on helping askers learn to distinguish the difference between the idea & how it's implemented, & to understand the ideas more thoroughly. $\endgroup$ Commented Oct 23, 2017 at 14:55
  • 5
    $\begingroup$ I am with @gung. It's a case of please post on Stack Overflow or a software-specific forum according to their criteria. (In fact, please debug my code is often a poor question even in those places.) $\endgroup$
    – Nick Cox
    Commented Oct 23, 2017 at 17:05
  • 6
    $\begingroup$ I'm not directly opposed to having a slightly more relaxed view of questions containing code than we presently do (we did in the past and I had no difficulty with that) as long as we don't start trying to compete with other sites that already cover it (so a pure-programming question should probably remain off topic). I am definitely still opposed to unclear questions (especially ones where you can't tell what they're even trying to do unless you can figure out what they wanted to do from the code). If we move this way we need to be able to spell out policy clearly in our help/on-topic. $\endgroup$
    – Glen_b
    Commented Oct 23, 2017 at 21:50
  • 4
    $\begingroup$ I agree with what's said above. Adding a personal note: the respectful scientific discourse of this forum is what attracted me, and has kept me as a contributor. I come from the more ML school of thought, with no rigorous statistical training, and this community has helped me become more well rounded. The consequence for this discussion is that, if the subject matter moves towards valuing programming as a core component of the skill-set (which I know, some long time posters would like to point out that it always has), then naturally new users will join that will help bring that... $\endgroup$ Commented Oct 24, 2017 at 0:07
  • 2
    $\begingroup$ ...perspective to our community. I would just encourage everyone to be welcoming when such valuable contributors appear. Possibly we should be discussing ways to offer mentorship to those that come through our doors. $\endgroup$ Commented Oct 24, 2017 at 0:08
  • 4
    $\begingroup$ Practicalities here as well as principles. 0. Stack Overflow exists! 1. The present boundary on software questions is a permanent problem likely to get worse if we get more indulgent. 2. Getting more indulgent will multiply questions without necessarily recruiting more people to answer. 3. Few things so unappealing as code questions on software you don't use. Nothing against software questions myself; I support SO and Statalist, but the present line on software questions although arbitrary and artificial is still better than that implied. $\endgroup$
    – Nick Cox
    Commented Oct 24, 2017 at 17:38
  • 4
    $\begingroup$ Least of all do I want on-line courses to outsource their student coding support to here, however unintended that may be! If the questions fit, that's fine. $\endgroup$
    – Nick Cox
    Commented Oct 24, 2017 at 17:39


You must log in to answer this question.

Browse other questions tagged .