- Reduce the extracted forest sizes upper bound and number because OMP seems to converge only with small forest sizes;
- Add extraction_strategy parameter in order to save base forest and the forests trained with the same size as the extracted forest sizes used in the experiment that used OMP.