Research map / Learning theory and algorithms / Prediction and decision losses
Gradient Boosted Decision Trees for High Dimensional Sparse Output
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 |
|---|---|---|---|
| Enforce L0 sparsity (k nonzeros) on leaf prediction vectors (ablation) | GBDT for high-dimensional sparse output · Wiki10-31K (Table 1) | Training time (s): 41078.76 -> 1054.12 s (about 40x) | Helped |
| Enforce L0 sparsity on leaf prediction vectors (ablation) | GBDT for high-dimensional sparse output · Wiki10-31K (Table 1) | Prediction time (s): 52.139 -> 1.087 s | Helped |
| Enforce L0 sparsity on leaf prediction vectors (ablation) | GBDT for high-dimensional sparse output · Wiki10-31K (Table 1) | P@1 (%): 84.11 -> 84.36 | No clear effect |
| Use LEML supervised projection instead of random projection for dense features (ablation) | GBDT-SPARSE · Wiki10-31K (Table 1) | P@1 (%): 80.79 -> 84.36 | Helped |
| Use PCA projection instead of LEML features (ablation) | GBDT-SPARSE · Wiki10-31K (Table 1) | P@1 (%): 84.36 -> 83.51 | Hurt |
| Use LEML projection (dimension reduction time 100.74 s) instead of random projection (4.97 s) (ablation) | GBDT-SPARSE · Wiki10-31K (Table 1) | Training time (s): 931.57 (random) -> 1054.12 (LEML) | Hurt |
| GBDT-SPARSE (sparse leaves and LEML features) vs. FASTXML | GBDT-SPARSE · Delicious-200K (Table 3) | Prediction time (s): 146.55 (FASTXML) vs 13.85 s | Helped |
| GBDT-SPARSE vs. FASTXML on Delicious-200K | GBDT-SPARSE · Delicious-200K (Table 3) | P@1 (%): 42.99 (FASTXML) vs 42.11 | Hurt |
| GBDT-SPARSE vs. FASTXML on Wiki10-31K | GBDT-SPARSE · Wiki10-31K (Table 3) | Model size (MB): 853.5 -> 85.81 MB | Helped |
| GBDT-SPARSE vs. SLEEC and PD-SPARSE on Wiki10-31K | GBDT-SPARSE · Wiki10-31K (Table 3) | P@1 (%): SLEEC 85.99, GBDT-SPARSE 84.34 | Hurt |
Closest papers by meaning
- Quantized Training of Gradient Boosting Decision Trees (NeurIPS 2022 · similarity 0.89)
- SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems (NeurIPS 2022 · similarity 0.89)
- Gradient Regularized Newton Boosting Trees with Global Convergence (NeurIPS 2026 · similarity 0.89)
- Navigating Extremes: Dynamic Sparsity in Large Output Spaces (NeurIPS 2024 · similarity 0.85)
- Near-Optimal Decision Trees in a SPLIT Second (ICML 2025 · similarity 0.85)
- Minimax Rates and Spectral Distillation for Tree Ensembles (NeurIPS 2026 · similarity 0.84)
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.