While we saw there is a lot of research indirectly concerned with ensemble learn-ing schemes, only little research is directly related to investigating Stacking, the
most general such scheme. A stronger research focus { outside the scope of
this thesis { seems to be on simpler ensemble learning schemes such as Bagging
(Breiman, 1996) and AdaBoost (F reund & Schapire, 1996), which combine only a
single kind of classier. While we believe that the current v ariants StackingC and
sMM5 are very close to the optimum
3
, we hope that our comprehensive overview
on related research stimulates further work in ensemble learning and applying
ensembles in other research areas such as Game Playing, Self-Diagnosing Sys-tems and BioInformatics