dmlc / dmlc/xgboost

Tutorial for interaction constraints is unclear about what the constraints apply to

Open
#10,462 0 comments 0 reactions 0 assignees View on GitHub
doc
Dominant language
C++
Stars
28.8k
Forks
8.9k
Avg merge
1d 12h
Merged PRs (30d)
54

Description

The docs for interaction contraints:
https://xgboost.readthedocs.io/en/stable/tutorials/feature_interaction_constraint.html

Are not very clear in describing what exactly are the constraints imposed by XGBoost, which seem to work differently than in other decision tree software.

Take for example the docs from H2O:
https://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/algo-params/interaction_constraints.html

It gives an example of the types of interactions that become forbidden and the ones that are still allowed, whereas in XGBoost's doc it's quite hard to understand and only mentioned towards the end that the constraints apply one step at a time only.

Contributor guide

No contributing guide indexed for this repository

Research direction

Start with the XGBoost interaction constraints tutorial linked in the issue and compare its explanation with the H2O documentation example. Clarify which interactions are forbidden or allowed when constraints are applied one step at a time, using concrete examples. Done means a reader can understand the constraint behavior without relying on the final brief mention.

Written by the indexing model from the issue text.

Assessment

Tech stack
machine-learning
Domain
documentation, machine-learning
Issue type
Documentation
Difficulty
2/5
Estimated time
1-3 hours
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
48/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.