There is data mining, which is like machine learning + databases. But for many, many machine learning problems, the learning occurs offline and the data is simp

There is data mining, which is like machine learning + databases. But for many, many machine learning problems, the learning occurs offline and the data is simply stored in files, then the learning algorithm is run. The result is a model, which in many cases has a small constant representation (common exceptions: kNN and SVM). There is no database involved. Now you can of course try to learn from an existing database (data mining) or from huge amounts of data, probably stored in a distributed database (big data). But many systems will not and even then the database part is usually not of much concern. So I still don't see much overlap.