Over time, a large number of machine learning questions have been asked and answered on Stack Overflow, amassing a significant amount of useful information. However, these questions are unfortunately off-topic for Stack Overflow, and as they come to members' attention, they are being closed (SO is for implementation questions, not theory). For example:
- What is the role of the bias in neural networks?
- What is the difference between supervised learning and unsupervised learning?
- Which machine learning classifier to choose, in general?
- What are advantages of Artificial Neural Networks over Support Vector Machines?
- Difference between classification and clustering in data mining?
- Why must a nonlinear activation function be used in a backpropagation neural network?
There are many more similar questions that are likely to be on the chopping block as mods get to them. And I'd really hate to lose the ability to have experts add new answers to them (or for them to be silently deleted if someone got overzealous without anyone noticing).
Unfortunately, there's a 60 day limit on migrating questions between sites. Moderators can't even override this, only SE employees can. This is for good reasons, smaller stacks didn't want StackOverflow dumping a backlog of terrible questions on small sites with even fewer moderation resources, or for bad questions to get passed around like White Elephants from stack to stack. Additionally, high-rep users on smaller stacks were rightfully upset that one migrated question from a big stack could shoot a member that never participated in their stack to the top of their rep ladders (and occasionally, vice versa).
This leaves SO in a difficult situation. Close the questions and let these questions stagnate? Delete them and hope the vacuum promotes similar questions to be asked on the correct stacks? Leave them open and set a bad precedent for newer theory questions (that should probably be asked here?)
I have two other possible solutions, but either would require the buy-in of the receiving stack. And I'm pretty sure the rightful home of most of these questions is here on Cross Validated.
Nominate a large block of well-maintained theory questions, link them here on stats.meta, get approval from CV members, then bring the list to se.meta and beg for migration.
- easier on mods, users, and google algorithms (which would be directed here instead of SO)
- high-rep questions getting migrated might distort the rep landscape of CV
- consensus on many questions is difficult to get simultaneously
- after all that effort, se.meta might just say no or leave it in limbo forever.
Manually copy "good" ML theory questions and answers from SO to CV when they are closed, and then ask for SO mods to add a redirect link here.
- Many of these old questions are asked/answered by users that may have not been active for years, so requesting reposts from them is likely a fool's errand.
- Manually "cross-posting" would be quite a lot of manual work, so it wouldn't be undertaken for anything but the most worthy Q&A's
- CV users can judge their worth for themselves and up/down/close vote as appropriate
- If I were to do this, I'd ask the mods to make both question and answer Community Wiki, because honestly I don't want to leech rep from work I didn't do. That is a controversial method, but personally, I'd rather err on the side of not taking credit for other's work.
I personally lean towards 2, although 1 would put most of the work on SE employee's shoulders (which the lazy slob in me likes) and bring more traffic here form Google (which may be a dubious benefit). But either would depend on buy-in from CV members and mods. So what do you think?
EDIT: After some delay, I've started closing questions on SO for this purpose. It would be helpful to have CV mods chime in on the MSO answer here for clarification on disputed questions (whether they are on-topic here, or may be on-topic elsewhere, like Data Science or AI)