There are 3 production sites: stackoverflow, stats.se, and math.se. There is some overflow between the sites, but at one extreme if your question is purely about an aspect of coding, use stackoverflow, and at the other extreme if your question is purely about an aspect of the maths, using maths.se. So for example, if you can't figure out the right way to organize your data prior to calling a regression library, use stackoverflow, but if your question is about how to use calculus to figure out a gradient in order to efficiently minimize your errors then use maths.se.
Stats.se (cross-validated) overlaps with both sites. Questions there often have some sort of programming element to them (especially in R or other statistical languages), but in the context of a statistical or machine learning problem. It's also a good place to ask question on topics including methodology, how to choose features, interpreting results, and the mathematics behind the algorithms.
All of these sites welcome questions from beginners, so long as they meet the sites' standards -- which in general means looking for solutions first especially on the site itself, describing a specific on-topic problem, and, where appropriate, the steps you have taken, or would take, to attempt to solve it. To get a good idea of the kinds of questions and answers on the different sites, you can look at the machine-learning tag on each site.
I'm not familiar with Computer Science, as mentioned in TGMClans' answer. But Browsing through the Computer Science questions tagged machine learning seems to confirm the description of Computer Science questions given in this answer
As a rule of thumb, if your question depends on code, ask on Stack Overflow; if your question calls for mathematical notation, ask on Computer Science. Algorithms expressed in pseudocode straddle the border.
You do, however, need to choose just one site for each question: don't cross-post. See here for more information.
Finally, here is a quantitative overview of the number of questions on the different sites.
Sources: SOSO, math, stats, compsci
Note that "unanswered" is a little misleading as it includes question which have answers with no votes, and also questions which are essentially answered in comments (this is particularly common on stats.se).
(Here is a related question on maths.se meta: Which site is better to ask question about Machine Learning?)
