Research map / Learning theory and algorithms / Prediction and decision losses
Counterfactual Explanations for Oblique Decision Trees:Exact, Efficient Algorithms
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 |
|---|---|---|---|
| Search the continuous feature space exactly instead of restricting candidates to training and test instances | Oblique tree counterfactuals · MNIST, Adult, Breast-Cancer, Spambase, Letter, German Credit | ℓ2 distance: MNIST: 53.50±17.24 -> 0.63±0.48; Adult: 3.4e4±1.2e5 -> 2.40±0.83; Breast: 0.86±0.49 -> 0.35±0.33; Spambase: 0.060±0.06 - | Helped |
| Use exact optimization instead of Feature Tweaking | Axis-aligned tree counterfactuals · MNIST, Breast-Cancer, Spambase, Letter, German Credit | ℓ2 distance: MNIST: 1.48±1.09 -> 0.04±0.11; Breast: 0.36±0.19 -> 0.13±0.17; Spambase: 0.005±0.009 -> 1.7e---5±0.0; Letter: 0.22±0.13 | Helped |
| Use exact optimization instead of Feature Tweaking | Axis-aligned tree counterfactuals · Adult | ℓ2 distance: 1.00±0.00 -> 2.05 ±0.31 | Hurt |
| Add user constraints on features | Oblique-tree counterfactuals · MNIST | ℓ2 distance: 0%: 0.63±0.48; 47%: 9.28±6.70 | Mixed |
| Add user constraints on features | Oblique-tree counterfactuals · Adult | ℓ2 distance: 0%: 2.40±0.83; 7%: 2.45±0.86; 14%: 2.49±0.97 | Mixed |
| Add user constraints | Training & test set search · MNIST | feasible counterfactuals (%): 100 -> 0 | Hurt |
| Add user constraints | Training & test set search · Breast-Cancer | feasible counterfactuals (%): 100 -> 17 | Hurt |
| Add user constraints | Training & test set search · Spambase | feasible counterfactuals (%): 100 -> 17 | Hurt |
| Add user constraints | Training & test set search · Letter | feasible counterfactuals (%): 100 -> 0 | Hurt |
| Add user constraints | Feature Tweaking · Adult | feasible counterfactuals (%): 28 | Hurt |
| Use ℓ1 rather than ℓ2 distance | Oblique-tree counterfactuals · MNIST | distance: ℓ2: 0.63±0.48; ℓ1: 2.85±1.40 | Mixed |
| Use a general quadratic distance with neighboring-pixel interactions | MNIST counterfactual images · MNIST | locality of changed pixels | Helped |
Closest papers by meaning
- Counterfactual Maps: What They Are and How to Find Them (NeurIPS 2026 · similarity 0.89)
- Generating Likely Counterfactuals Using Sum-Product Networks (ICLR 2025 · similarity 0.88)
- A General Search-Based Framework for Generating Textual Counterfactual Explanations (AAAI 2024 · similarity 0.88)
- Promoting Counterfactual Robustness through Diversity (AAAI 2024 · similarity 0.88)
- CF-OPT: Counterfactual Explanations for Structured Prediction (ICML 2024 · similarity 0.87)
- DiCoFlex: Model-Agnostic Diverse Counterfactuals with Flexible Control (NeurIPS 2025 · similarity 0.87)
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.