Research map / Adversarial robustness and security / Certified and empirical robustness
Abstract Interpretation of Decision Tree Ensemble Classifiers
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 silva instead of VoTE | tree ensemble verifier · MNIST GBDT 100x10 | verification time (s): 974 -> 147 | Helped |
| use silva instead of VoTE | tree ensemble verifier · MNIST RF 50x10 | verification time (s): 7704 -> 4353 | Helped |
| use silva instead of VoTE | tree ensemble verifier · MNIST RF 50x10 | robustness (%): 31.4 -> 30.2 | No clear effect |
| average vote instead of max vote in random forest (ablation) | RF entropy 5x5 · MNIST | stability (%): 9.6 -> 20.2 | Helped |
| average vote instead of max vote in large forest (ablation) | RF entropy 50x25 · MNIST | stability (%) | No clear effect |
| train with entropy instead of Gini criterion (ablation) | RF 50x25 average · MNIST | stability (%): 35.2 -> 43.1 | Helped |
| increase max tree depth from 25 to 50 (ablation) | RF 25 trees entropy average · MNIST | stability (%): 39.4 -> 36.4 | Hurt |
| increase max tree depth from 25 to 50 (ablation) | RF 5 trees entropy average · MNIST | stability (%): 28.4 -> 22.8 | Hurt |
| increase number of trees from 5 to 50 at depth 25 (ablation) | RF entropy average · MNIST | stability (%): 28.4 -> 43.1 | Helped |
| increase depth of RFs from 5 to 20 (ablation) | RF 20 trees · Sensorless | stability (%): 27.9 -> 17.9 | Hurt |
| increase GBDT depth from 6 to 7 (ablation) | GBDT 25 trees · Sensorless | stability (%): 22.8 -> 14.8 | Hurt |
| increase number of GBDT trees from 25 to 30 (ablation) | GBDT depth 6 · Sensorless | stability (%): 22.8 -> 20.2 | Hurt |
Closest papers by meaning
- Robustness Verification of Tree-based Models (NeurIPS 2019 · similarity 0.92)
- Robustness Verification of Multi-Class Tree Ensembles (AAAI 2024 · similarity 0.90)
- Sensitivity Verification for Additive Decision Tree Ensembles (ICLR 2025 · similarity 0.90)
- Multistage Defer Trees for Hybrid Interpretability: If at First You Can't Succeed, Tree Again (NeurIPS 2026 · similarity 0.84)
- Learning Decision Trees and Forests with Algorithmic Recourse (ICML 2024 · similarity 0.84)
- Probably Approximately Global Robustness Certification (ICML 2025 · 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.