this doesn't work in all cases, e.g. when doing bagging, but often it is what one wants: get the last PipeOp in a sequential Graph that is a PipeOpLearner, and return its (trained) Learner.
maybe also just forward all methods / slots of the last learner to GraphLearner. This is anarchy, but it would solve quite a few Tag: Graph Transparency problems.
this doesn't work in all cases, e.g. when doing bagging, but often it is what one wants: get the last
PipeOp
in a sequentialGraph
that is aPipeOpLearner
, and return its (trained)Learner
.