Outsourcing HO and implementing it also for BDTs
This MR
- outsources the HO into a dedicated file
- implements HO for BDTs
- Allows for
LearningRate
to be HO parameter everywhere; steered byMaxLearningRate
,MinLearningRate
andStepLearningRate
- Allows for
MaxDepth
andMaxFeatures
to be HO parameters for HO for BDTs. The code will take the given values as maxima and also produce shallower (smallerMaxDepths
) models with lessMaxFeatures
. User should therefore consider to increase these values for HO to receive a wider range of configurations.