Extending non-ensemble learners, e.g. C4.5, to incorporate more and different
biases. In a way this is complementary to Stacking: whereas Stacking
combines the output from its component classifiers without making
any assumptions about their internal structure, the mentioned approach
focusses on meshing classifiers to hopefully combine their strengths and
mitigate their weaknesses. So we also consider these extended learners to
be – somewhat distant – relatives to Stacking