Custom Constraints for XGBoost Training

Is there any way to have custom constraints specified during XGBoost training, particularly when training with sample weights? For example, for my use case, I want to have constraints like:

  • constraints on all the samples on the sample weights (e.g. no leaf should have too much sample weight based on the samples in that leaf)
  • constraints on individual samples (e.g. no individual sample should ever end up in a leaf by itself)