This is an exploratory workstream. We have observed the GFR algorithm to yield odd / unstable results when features have many ties. We would like to better understand what makes this the case and determine whether any workarounds exist.
In service of this goal, we'd also like to work out a principled approach to handling differences in the number of available cutpoints across features (see #167).
This is an exploratory workstream. We have observed the GFR algorithm to yield odd / unstable results when features have many ties. We would like to better understand what makes this the case and determine whether any workarounds exist.
In service of this goal, we'd also like to work out a principled approach to handling differences in the number of available cutpoints across features (see #167).