Declarative Recursive Computation on an RDBMS

We explore the close relationship between the tensor-based computations performed during modern machine learning, and relational database computations. We consider how to make a very small set of changes to a modern RDBMS to make it suitable for distributed learning computations. Changes include adding better support for recursion, and optimization and execution of very large compute plans. We also show that there are key advantages to using an RDBMS as a machine learning platform. In particular, DBMS-based learning allows for trivial scaling to large data sets and especially large models, where different computational units operate on different parts of a model that may be too large to fit into RAM.