Can feature_weights can be used together with monotone_constraints?

#1

I have enforced monotone_constraints on a feature, however I noticed that feature importance of that feature dropped dramatically.
I would like to increase weight of it, and set weight with dtrain.set_info(feature_weights=fw), however it seems that it has no effect, when monotone_constraints is set.

My question:
Can I make the two work together?

Best Regards,
DK

#2

The weight affects column sampling. If a feature violates mono constraints then it will be discarded even if it’s sampled.

#3

OK, thank you for the clarification!

DK