* learning.accuracy01 should embed a computable core (which can then be coerced to R) * empiricalloss.num_correct should call the computable core of learning.accuracy01
