13

Proposal: Machine Learning

It seems that there was already a machine learning Q&A site on this platform and that it didn't work too well. It was merged with CrossValidated, the stats Q&A site. What's the problem with this solution?

9

The difference is that we try again. :-)

Many times things like this are 'non-linear' so to speak. You have to keep trying and keep trying to drum up interest until it succeeds. I think ML is growing and it is just a matter of time.

  • 5
    But what's the point? Why keep trying if those question are properly answered on the current site? – Gala Jan 29 '13 at 9:15
  • 10
    Well, no, that's simply not true. Pure stat questions with no coding aspect weren't really appropriate for stackoverflow and wouldn't have reach the right people. – Gala Jan 30 '13 at 13:51
  • 3
    But then, if you think that this doesn't matter, why not a site on random forests then? SVM? Neural nets? The GLM? Or perhaps one for supervised learning/predictive modeling and another one for unsupervised learning/clustering (this division runs through both stat and ML)? – Gala Jan 30 '13 at 13:53
  • 1
    It seems to me that you would need some reason to create more specialized sites (perhaps the more general one is too big or does not attract the right people…) but I fail to see one in this case. – Gala Jan 30 '13 at 13:54
  • 4
    At the end of the day, there is so much overlap between statistics and ML that this all seems like a complete waste of time and resources. Until there is a compelling reason to split the community, I know that as a user I would rather be able to find all relevant material and competent people in one place. – Gala Jan 30 '13 at 13:56
  • 4
    I also believe that instead of identifying strongly with stats/ML/data mining/predictive analytics/data science, most people would benefit a lot from learning a little more about these fields as they have a lot in common so there is a good reason to keep all these topics together as long as possible. – Gala Jan 30 '13 at 13:58
  • 2
    That's why I am asking what the problem is with the current solution. “We have to try until it succeeds” or “We need to mark our territory” do not seem like good answers to this question. – Gala Jan 30 '13 at 14:01
  • 3
    @GaëlLaurans I believe the short answer to your question is that ML is a field unto itself in this day and age. You can get a PhD in ML. I think fields like this do have overlap, but ML exists as a field unto itself. If you are asking "why is a field a field?", then the short answer is "Because it just is.". Empirically ML has become its own field, so I believe it warrants its own site. – Spacey Jan 30 '13 at 15:56
  • 2
    I guess what I am trying to get at are the practical advantages of separate sites for ML and statistics. I don't feel that “because ML is a field, period” is a relevant answer at all. – Gala Jan 31 '13 at 17:13
  • 1
    How would ML students or practitioners concretely benefit from a new site (which would obviously have to start without existing community or backlog of relevant questions)? In what way are they not well served by the stat site? Given the previous failure and the drawbacks of having relevant material and interested people spread over two different sites, I think you need some serious, actual problems with the current solution before wasting energy trying to get a new community off the ground. – Gala Jan 31 '13 at 17:17
  • 2
    The questions I asked were (1) what makes you think it's going to work this time and (2) what's the problem with crossvalidated covering both stat and ML? Instead of addressing them, you just assume that creating a separate site is an unquestionable goal and that it should be tried again and again no matter what. I am sorry but stating “it's a field” isn't an explanation of anything, let alone a concrete one. You're completely free to ignore me, to try anything you want, or to think that my questions are not important but you certainly haven't answered them. – Gala Feb 2 '13 at 12:30
  • 9
    @Mohammad What I would be interested in, though, is: Have you tried to use CrossValidated? – stacky-bit Feb 3 '13 at 14:55
  • 1
    Well, I never spoke of a practical site but an example of a practical problem would be the absence of an appropriate place to ask a particular set of questions. This is basically the problem that many new sites address by opening up the StackExchange universe to new topics. But that would not be the case here, for ML is already covered by CrossValidated. Hence my asking: What's the problem? – Gala Feb 7 '13 at 23:10
  • 2
    Yet you spent more than a week discussing with me while conscientiously eluding the question I asked. It seems you don't want to let it rest so let me try to explain it again in a different way. Basically, what I am asking is: What would be achieved by this new site? What reason do we have to think it can succeed so shortly after a failure? What practical benefits would it bring to readers and contributors? etc. – Gala Feb 13 '13 at 10:21
  • 4
    I explained in the comments why I think CrossValidated is a good solution and I still think it would be a good idea to lay out some practical reasons to set up a separate site. Your answer above and most of your comments just assume that it is a valuable goal but never explain why, especially not on a practical level. – Gala Feb 13 '13 at 10:25

You must log in to answer this question.

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