We have two relatively popular tags lg.learning (53 questions) and machine-learning (75 questions). However, I am not sure where the line between the two is drawn. For lg.learning wiki we have:
Tag lg.learning refers to the research area of machine learning, and learning theory specifically (falling under arXiv's cs.LG - Learning), as opposed to the general practice of learning. This tag includes the theory of PAC learning, algorithmic learning theory, and computational aspects of Bayesian inference and graphical models. Questions about implementation issues and statistical properties of machine learning systems are more likely to be welcomed at the CrossValidated or MetaOptimize Q&A sites.
The questions must satisfy the usual scope requirements for cstheory as explained in the FAQ.
And for machine-learning we have the wiki-excerpt:
Theoretical questions about Machine learning, especially Computational Learning Theory, including Algorithmic Learning Theory, PAC learning, and Bayesian Inference
and the whole wiki:
This tag should be used for theoretical questions related to Machine Learning (roughly falling under arXiv's cs.LG - Learning).
The question must satisfy the usual scope requirements for cstheory as explained in the FAQ.
You may also want to check AI Q&A site Meta Optimize which has a more general AI scope.
To me, it seems that we are using them as synonyms. Is this the case? If so, then which should be the main tag? lg.learning follows our ArXiv tag convention, but machine-learning is more popular and more likely to be searched for.