Yu-Group / imodels-experiments

Experiments with experimental rule-based models to go along with imodels.
https://csinva.io/imodels/
15 stars 4 forks source link

CCP and number of leaves #17

Open markusloecher opened 8 months ago

markusloecher commented 8 months ago

In Fig. 4A/B the CCP (cost complexity pruned) graphs are puzzling to me: CCP should protect against overfitting a lot more than shown? In particular, the "number of leaves" cannot be fixed as the graph suggests; they will be a result of the pruning?