Feature importances as calculated during the model training and
[0; 1]. The normalization is done by dividing
the accumulated feature importances over the maximum value.
A frame with two columns:
feature_name that has stype
feature_importance that has stype
depending on whether the