Research map / Learning theory and algorithms / Sample complexity
Fast Rates for a kNN Classifier Robust to Unknown Asymmetric Label Noise
Henry Reeve, Ata Kaban
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 |
|---|---|---|---|
| Estimate corrupted regression and noise rates from kNN neighborhood averages and empirical extrema, then correct the classification threshold | Robust kNN · Asymmetric label noise with unknown noise probabilities | excess risk: \left( \frac{\log(18n/\delta)}{n}\right)^{\frac{\lambda(\alpha+1)}{2\lambda+1}} +\delta | Improved |
| Use standard kNN without correcting for asymmetric label noise | kNN · Asymmetric label noise | consistency | Hurt |
| Estimate a function's maximum by maximizing kNN regression estimates at sample points | kNN empirical maximum estimator · Measure-smooth function estimation | maximum estimation error: \sqrt{\frac{\log(6n/\delta)}{2k}}+2\omega \cdot \left(\frac{2k}{n}\right)^{\lambda} | Improved |
| Train a classifier consistent for the corrupted distribution, then test it on the clean distribution when their regression functions disagree across the decision boundary | Any classically Bayes-consistent classifier · Asymmetric label noise with positive-measure disagreement region | consistency on the clean test distribution: \theta \cdot \mu\left(\mathcal{A}_{\theta}(\eta,\noisy[\eta])\right)>0 | Hurt |
Closest papers by meaning
- Robustness of Accuracy Metric and its Inspirations in Learning with Noisy Labels (AAAI 2021 · similarity 0.88)
- Leveraging Data Symmetries to Select an Optimal Subset of Training Data under Label Noise (NeurIPS 2026 · similarity 0.88)
- Efficient Classification with Adaptive KNN (AAAI 2021 · similarity 0.88)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.87)
- Information-theoretic Limits of Online Classification with Noisy Labels (NeurIPS 2024 · similarity 0.86)
- Estimation of the Label-Noise Transition Matrix with Performance Guarantees via Selective Classification (NeurIPS 2026 · 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.
