Research map / Learning theory and algorithms / Prediction and decision losses
From global to local MDI variable importances for random forests and when they are Shapley values
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 randomization parameter K of Extra-Trees | global MDI and SAGE · led | normalized importance | Mixed |
| compare global MDI with SAGE (cross-entropy loss) at K=1 | global MDI · led | importance scores | No clear effect |
| compare SAGE with mean squared loss and TreeSHAP perturbation options | SAGE and TreeSHAP · led | importance vs K | No clear effect |
| compare local MDI with Saabas and TreeSHAP | local MDI · digits | correlation of local importance scores | No clear effect |
| compare local MDI with Saabas, Pearson | local MDI · Wine | fraction of samples with correlation >= 0.9: 72.47 | Mixed |
| compare local MDI with TreeSHAP, Pearson | local MDI · Digits | fraction of samples with correlation >= 0.9: 38.01 | Mixed |
| compare local MDI with Saabas, Pearson | local MDI · Led | fraction of samples with correlation >= 0.9: 100.00 | No clear effect |
Closest papers by meaning
- A Debiased MDI Feature Importance Measure for Random Forests (NeurIPS 2019 · similarity 0.89)
- Conditional Feature Importance with Generative Modeling Using Adversarial Random Forests (AAAI 2025 · similarity 0.85)
- Regional Explanations: Bridging Local and Global Variable Importance (NeurIPS 2025 · similarity 0.85)
- A Unified Framework for Provably Efficient Algorithms to Estimate Shapley Values (NeurIPS 2025 · similarity 0.84)
- Accurate Estimation of Feature Importance Faithfulness for Tree Models (AAAI 2025 · similarity 0.83)
- Feature Importance Metrics in the Presence of Missing Data (ICML 2025 · similarity 0.83)
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.