Research map / Learning theory and algorithms / Prediction and decision losses
LightGBM: A Highly Efficient Gradient Boosting Decision Tree
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add GOSS and EFB (LightGBM) vs. lgb_baseline | LightGBM · Allstate, Flight Delay, LETOR, KDD10, KDD12 | training time per iteration: speed-up of 21x, 6x, 1.6x, 14x, 13x vs lgb_baseline | Helped |
| Add GOSS and EFB | LightGBM · Allstate, Flight Delay, LETOR, KDD10, KDD12 | test accuracy (AUC / NDCG@10): differences within about 0.0002 AUC/NDCG of lgb_baseline | No clear effect |
| Add GOSS (on top of EFB) (ablation) | LightGBM · Allstate, Flight Delay, LETOR, KDD10, KDD12 | training time per iteration: nearly 2x speed-up by GOSS alone (EFB_only vs LightGBM) | Helped |
| Add EFB (EFB_only vs lgb_baseline) (ablation) | lgb_baseline · Allstate, KDD10, KDD12, Flight Delay, LETOR | training time per iteration: e.g. Allstate 6.07 to 0.71 s; KDD10 39.85 to 6.33 s | Helped |
| Use GOSS instead of stochastic gradient boosting (SGB) at same sampling ratio (ablation) | LightGBM sampling on LETOR · LETOR NDCG@10 across sampling ratios 0.1-0.4 | NDCG@10: e.g. 0.5224 vs 0.5182 at ratio 0.1; 0.5296 vs 0.5267 at ratio 0.4 | Helped |
| Use stochastic gradient boosting sampling instead of full data (SGB) (ablation) | lgb_baseline · Allstate, Flight Delay, LETOR, KDD10, KDD12 | test accuracy (AUC / NDCG@10): e.g. LETOR 0.5239 vs 0.5277; KDD12 0.6989 vs 0.7049 | Hurt |
| Use histogram-based xgb_his instead of pre-sorted xgb_exa | XGBoost · KDD10, KDD12 | memory / runnability: OOM on KDD10 and KDD12 | Hurt |
Closest papers by meaning
- Quantized Training of Gradient Boosting Decision Trees (NeurIPS 2022 · similarity 0.91)
- Gradient Regularized Newton Boosting Trees with Global Convergence (NeurIPS 2026 · similarity 0.87)
- NRGBoost: Energy-Based Generative Boosted Trees (ICLR 2025 · similarity 0.87)
- GradTree: Learning Axis-Aligned Decision Trees with Gradient Descent (AAAI 2024 · similarity 0.86)
- Near-Optimal Decision Trees in a SPLIT Second (ICML 2025 · similarity 0.85)
- Symbolic Regression Enhanced Decision Trees for Classification Tasks (AAAI 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.