Welcome to wideboost!

wideboost implements Wide Boosting as described in this article. wideboost does this by wrapping existing popular Gradient Boosting packages (XGBoost and LightGBM). If you're familiar with those packages wideboost can essentially be used as a drop-in replacement for either XGBoost or LightGBM when you're using one of the wideboost supported objective functions.

Wide Boosting tweaks the usual Gradient Boosting framework as described in our Overview. It solves the same problems as Gradient Boosting. On several datasets is exhibits much better performance (see the Overview or the article). Supported objective functions for wideboost include usual univariate and multivariate regression and binary and multi-category classification.

Since Wide Boosting is closely related to Gradient Boosting, we can use the same tools to interpret a wideboost model. wideboost includes a wrapper on SHAP to aid in interpreting a wideboost model.

• Installation
• Overview
• Examples
• Supported objective functions
• XGBoost wrapper
• LightGBM wrapper
• SHAP explainer

Reference

Horrell, M. (2020). Wide Boosting. arXiv preprint arXiv:2007.09855.