Flexible model composition in machine learning and its implementation in MLJ

A graph-based protocol called `learning networks' which combine assorted\nmachine learning models into meta-models is described. Learning networks are\nshown to overcome several limitations of model composition as implemented in\nthe dominant machine learning platforms. After illustrating the protocol in\nsimple examples, a concise syntax for specifying a learning network,\nimplemented in the MLJ framework, is presented. Using the syntax, it is shown\nthat learning networks are are sufficiently flexible to include Wolpert's model\nstacking, with out-of-sample predictions for the base learners.\n