Research map / Learning theory and algorithms / Prediction and decision losses
Statistical Analysis of Nearest Neighbor Methods for Anomaly Detection
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 IForest rather than kNN | Unsupervised anomaly detection on synthetic benchmarks · Synthetic benchmark corpus | Either-metric failure rate: 0.6960 -> 0.6741 | Helped |
| Use kNN rather than IForest | High-dimensional anomaly detection · gisette, isolet, letter | AUC: 0.5023 -> 0.5696; 0.5485 -> 0.6810; 0.5600 -> 0.7162 | Helped |
| Use LOF rather than IForest | High-dimensional anomaly detection · gisette, isolet, letter | AUC: 0.5023 -> 0.6753; 0.5485 -> 0.7330; 0.5600 -> 0.7846 | Helped |
| Use kNN rather than LOF | Anomaly detection on moderately lower-dimensional datasets · cancer, ionosphere | AUC: 0.8097 -> 0.9780; 0.9450 -> 0.9832 | Helped |
| Compare the methods on 23 ODDS datasets | IForest, LODA, LOF, DTM₂, kNN, and kᵗʰNN · ODDS library | AUC and AP | No clear effect |
| Replace global NN scoring with LOF-style local scoring | Anomaly detection for locally isolated anomalies · Illustrative local-anomaly example | Anomaly detection performance | Helped |
| Use LOF rather than NN methods for clustered global anomalies | Anomaly detection for clustered anomalies · Illustrative clustered-anomaly example | Anomaly detection performance | Hurt |
| Reduce separation between anomalies and normal observations | DTM₂ anomaly scoring · Illustrative separation example | Misclassification rate | Hurt |
Closest papers by meaning
- Quantifying Statistical Significance of Deep Nearest Neighbor Anomaly Detection via Selective Inference (NeurIPS 2025 · similarity 0.88)
- Deep Semi-Supervised Anomaly Detection (ICLR 2020 · similarity 0.87)
- LUNAR: Unifying Local Outlier Detection Methods via Graph Neural Networks (AAAI 2022 · similarity 0.86)
- A Theory-Based Evaluation of Nearest Neighbor Models Put Into Practice (NeurIPS 2018 · similarity 0.86)
- Prototypical Residual Networks for Anomaly Detection and Localization (CVPR 2023 · similarity 0.86)
- Towards a Unified Framework of Clustering-based Anomaly Detection (ICML 2025 · 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.