| # UltraGBM: Unified GBDT Library Research |
|
|
| Comprehensive deep research covering ALL algorithmic advances in Gradient Boosted Decision Tree families. |
|
|
| **Date:** 2026-04-29 |
|
|
| ## Covered Frameworks |
| 1. **XGBoost** (Chen & Guestrin, KDD 2016) — arxiv:1603.02754 |
| 2. **LightGBM** (Ke et al., NeurIPS 2017) |
| 3. **CatBoost** (Prokhorenkova et al., 2017/2018) — arxiv:1706.09516 |
| 4. **Yggdrasil/TF-DF** (Guillame-Bert et al., KDD 2023) — arxiv:2212.02934 |
| 5. **Recent Advances** (2022-2026): GRANDE, NRGBoost, GBRL, NODE, C-GB, DP-EBM, HybridTree |
|
|
| ## Key Innovations Catalogued |
| - 22 algorithmic innovations ranked by priority |
| - Exact equations, pseudocode, and hyperparameters for each |
| - Master recipe table with implementation complexity vs. impact |
| - Phased architecture roadmap for UltraGBM implementation |
|
|
| See `UltraGBM_Research_Report.md` for the full document. |
