Research map / Adversarial robustness and security / Certified and empirical robustness
Robust Optimal Classification Trees against Adversarial Examples
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 |
|---|---|---|---|
| train robust trees with MILP and a GROOT warm start | ROCT · 8 datasets (aggregate) | mean adversarial accuracy: .726 -> .735 | Helped |
| add a GROOT warm start to continuous-threshold MILP (ablation) | ROCT MILP · 8 datasets (aggregate) | mean adversarial accuracy: .720 -> .735 | Helped |
| add a GROOT warm start to binary-threshold MILP (ablation) | ROCT Binary-MILP · 8 datasets (aggregate) | mean adversarial accuracy: .714 -> .726 | Helped |
| use continuous rather than binary threshold variables, with GROOT initialization (ablation) | ROCT · 8 datasets (aggregate) | mean adversarial accuracy: .726 -> .735 | Helped |
| use continuous rather than binary threshold variables without warm start (ablation) | ROCT · 8 datasets (aggregate) | mean adversarial accuracy: .714 -> .720 | Helped |
| use LSU-MaxSAT rather than MILP-warm (ablation) | ROCT · 8 datasets (aggregate) | mean adversarial accuracy: .735 -> .729 | Hurt |
| use TREANT rather than MILP-warm | robust decision tree training · 8 datasets (aggregate) | mean adversarial accuracy: .735 -> .692 | Hurt |
| use LSU-MaxSAT during early optimization rather than MILP-warm | depth-3 ROCT training · 8 datasets (mean over datasets) | training misclassification percentage: approximately 10 and 100 seconds; approximately 1000 seconds; 30 minutes | Mixed |
| use GROOT rather than an optimal LSU-MaxSAT tree | robust decision trees · training datasets, depth up to 2 | training adversarial accuracy ratio: 0.92; approximately 0.87 | No clear effect |
Closest papers by meaning
- Robust Decision Trees Against Adversarial Examples (ICML 2019 · similarity 0.92)
- Efficient Training of Robust Decision Trees Against Adversarial Examples (ICML 2021 · similarity 0.89)
- TTTS: Tree Test Time Simulation for Enhancing Decision Tree Robustness against Adversarial Examples (AAAI 2024 · similarity 0.87)
- Coevolutionary Algorithm for Building Robust Decision Trees under Minimax Regret (AAAI 2024 · similarity 0.86)
- Faster Repeated Evasion Attacks in Tree Ensembles (NeurIPS 2024 · similarity 0.86)
- Robustness Verification of Multi-Class Tree Ensembles (AAAI 2024 · similarity 0.85)
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.