Research map / Learning theory and algorithms / Prediction and decision losses
Finding Influential Training Samples for Gradient Boosted Decision Trees
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 |
|---|---|---|---|
| Increase the FastLeafRefit update set size (ablation) | FastLeafRefit · Adult, Same | NDCG@100: 0.38 -> 1.00 | Helped |
| Increase the FastLeafRefit update set size (ablation) | FastLeafRefit · Adult, Changed | NDCG@100: 0.10 -> 0.10 | No clear effect |
| Increase the FastLeafInfluence update set size (ablation) | FastLeafInfluence · Adult, Same | NDCG@100: 0.39 -> 1.00 | Helped |
| Increase the FastLeafInfluence update set size (ablation) | FastLeafInfluence · Adult, Changed | NDCG@100: 0.80 -> 1.00 | Helped |
| Use LeafInfluence rather than LeafRefit when the tree structure assumption is violated | GBDT influence estimation · Adult, Changed | NDCG@100 | Helped |
| Use a smaller update set (ablation) | FastLeafRefit and FastLeafInfluence · Runtime comparison | runtime | Helped |
| Use FastLeafInfluence instead of FastLeafRefit | GBDT influence estimation · Runtime comparison | runtime | Helped |
| Use small update sets to target individual test cases (ablation) | FastLeafRefit and FastLeafInfluence · Label-noise removal | whole-test-set Logloss reduction | Hurt |
| Rank training examples by influence for a specific test object | FastLeafRefit and FastLeafInfluence · Label-noise removal | DCG of relative Logloss reductions | Helped |
| Use influence estimates to identify examples from the underrepresented age group with positive labels | FastLeafRefit and FastLeafInfluence · Hospital domain mismatch, age∈[40;50) | average test Logloss influence | Helped |
Closest papers by meaning
- Sensitivity Verification for Additive Decision Tree Ensembles (ICLR 2025 · similarity 0.90)
- Pushing the Envelope of Gradient Boosting Forests via Globally-Optimized Oblique Trees (CVPR 2022 · similarity 0.88)
- On Computing Optimal Tree Ensembles (ICML 2023 · similarity 0.88)
- Multistage Defer Trees for Hybrid Interpretability: If at First You Can't Succeed, Tree Again (NeurIPS 2026 · similarity 0.86)
- GRANDE: Gradient-Based Decision Tree Ensembles for Tabular Data (ICLR 2024 · similarity 0.86)
- Minimax Rates and Spectral Distillation for Tree Ensembles (NeurIPS 2026 · similarity 0.86)
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.