Research map / Learning theory and algorithms / Prediction and decision losses
SketchBoost: Fast Gradient Boosted Decision Tree for Multioutput Problems
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 |
|---|---|---|---|
| Use Random Projection sketching for split search (ablation) | SketchBoost · Dionis | cross-entropy: 0.3114 -> 0.2848 | Helped |
| Use Random Projection sketching for split search (ablation) | SketchBoost · Otto | cross-entropy: 0.4697 -> 0.4566 | Helped |
| Use Random Sampling sketching for split search (ablation) | SketchBoost · SCM20D | RMSE: 91.0142 -> 86.2964 | Helped |
| Use Top Outputs sketching for split search (ablation) | SketchBoost · Dionis | cross-entropy: 0.3114 -> 0.3146 | No clear effect |
| Use Random Projection rather than full-output split scoring (ablation) | SketchBoost · Dionis | training time per fold: 23919 -> 419 | Helped |
| Reduce sketch dimension k (ablation) | SketchBoost sketching strategies · Multiple datasets | early-iteration validation error decay | Hurt |
| Increase sketch dimension k (ablation) | SketchBoost sketching strategies · Multiple datasets | test error | Helped |
| Use Random Projection instead of SketchBoost Full (ablation) | SketchBoost · Dionis | training time per fold: 23919 -> 419 | Helped |
| Use a single-tree strategy rather than one-versus-all | Multioutput GBDT · Most datasets | generalization performance | Helped |
| Use SketchBoost rather than CatBoost | Multitask GBDT · RF1 | RMSE: 0.8975 -> 0.9056 | Hurt |
| Use SketchBoost rather than CatBoost | Multitask GBDT · SCM20D | RMSE: 90.9814 -> 85.8061 | Helped |
| Use Random Projection rather than SketchBoost Full (ablation) | SketchBoost · SF-Crime | cross-entropy: 2.2067 -> 2.2038 | No clear effect |
Closest papers by meaning
- Quantized Training of Gradient Boosting Decision Trees (NeurIPS 2022 · similarity 0.90)
- LightGBM: A Highly Efficient Gradient Boosting Decision Tree (NeurIPS 2017 · similarity 0.89)
- NRGBoost: Energy-Based Generative Boosted Trees (ICLR 2025 · similarity 0.85)
- Gradient Regularized Newton Boosting Trees with Global Convergence (NeurIPS 2026 · similarity 0.85)
- Can Gaussian Sketching Converge Faster on a Preconditioned Landscape? (ICML 2024 · similarity 0.84)
- DOGE-Train: Discrete Optimization on GPU with End-to-End Training (AAAI 2024 · 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.