FIGS classifier - leafs with small number of instances

[filipr]

Hi,
first of all thanks for your work - I have found both the article and the code for FIGS very interesting.

I have recently tried FIGS models for some classification task with quite unbalanced dataset, and quite often it happens that I see a subtree where only a few observations (the dataset contains ~100k rows and in some subtree there would be only 5-10 instances).

Is there a way how to limit such behavior (potentially something like min_samples_split parameter)?