Research map / Learning theory and algorithms / Prediction and decision losses
Optimal Decision Tree with Noisy Outcomes
Su Jia, viswanath nagarajan, Fatemeh Navidi, R Ravi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| ODTNr (adaptive, O(r+log m)) instead of Non-Adap | ODT with noisy outcomes · WISER (unique stopping) | expected number of tests: 11.568 -> 8.357 | Improved |
| ODTNr instead of Non-Adap | ODT with noisy outcomes · CL-30 | expected number of tests: 10.204 -> 8.000 | Improved |
| ODTNr instead of ODTNh | ODT with noisy outcomes · CL-30 | expected number of tests: 8.729 -> 8.000 | Improved |
| ODTNr instead of ODTNh | ODT with noisy outcomes · WISER (unique stopping) | expected number of tests: 9.707 -> 8.357 | Improved |
| ODTNr vs ODTNh with no noise (CL-0) | ODT with noisy outcomes · CL-0 | expected number of tests: 7.910 -> 7.910 | No clear effect |
| ODTNr instead of Non-Adap under neighborhood stopping | ODT with noisy outcomes · Original WISER, neighborhood stopping | expected number of tests: 16.995 -> 11.163 | Improved |
| ODTNr instead of Non-Adap under clique stopping | ODT with noisy outcomes · Original WISER, clique stopping | expected number of tests: 21.281 -> 11.817 | Improved |
| ODTNr instead of ODTNh under clique stopping | ODT with noisy outcomes · Original WISER, clique stopping | expected number of tests: 12.506 -> 11.817 | Improved |
| skip non-informative tests adaptively (Low-Adap vs Non-Adap) | Non-Adap · WISER (unique stopping) | expected number of tests: 11.568 -> 9.152 | Improved |
| Low-Adap vs Non-Adap with neighborhood stopping | Non-Adap · Original WISER, neighborhood stopping | expected number of tests: 16.995 -> 16.983 | No clear effect |
| increase classifier margin noise threshold d from 0 to 30 | ODTNh · CL-d | expected number of tests: 7.910 -> 8.729 | Hurt |
| increase classifier margin noise threshold d from 0 to 30 | ODTNr · CL-d | expected number of tests: 7.910 -> 8.000 | No clear effect |
Closest papers by meaning
- MAPTree: Beating “Optimal” Decision Trees with Bayesian Decision Trees (AAAI 2024 · similarity 0.89)
- Popular decision tree algorithms are provably noise tolerant (ICML 2022 · similarity 0.88)
- Active Learning for Decision Trees with Provable Guarantees (ICLR 2026 · similarity 0.88)
- Robust Loss Functions for Training Decision Trees with Noisy Labels (AAAI 2024 · similarity 0.87)
- Improving Decision Trees through the Lens of Parameterized Local Search (NeurIPS 2025 · similarity 0.86)
- Fully-Dynamic Approximate Decision Trees With Worst-Case Update Time Guarantees (ICML 2024 · 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.
