English

Learning Efficient Anomaly Detectors from $K$-NN Graphs

Machine Learning 2015-02-09 v1 Machine Learning

Abstract

We propose a non-parametric anomaly detection algorithm for high dimensional data. We score each datapoint by its average KK-NN distance, and rank them accordingly. We then train limited complexity models to imitate these scores based on the max-margin learning-to-rank framework. A test-point is declared as an anomaly at α\alpha-false alarm level if the predicted score is in the α\alpha-percentile. The resulting anomaly detector is shown to be asymptotically optimal in that for any false alarm rate α\alpha, its decision region converges to the α\alpha-percentile minimum volume level set of the unknown underlying density. In addition, we test both the statistical performance and computational efficiency of our algorithm on a number of synthetic and real-data experiments. Our results demonstrate the superiority of our algorithm over existing KK-NN based anomaly detection algorithms, with significant computational savings.

Keywords

Cite

@article{arxiv.1502.01783,
  title  = {Learning Efficient Anomaly Detectors from $K$-NN Graphs},
  author = {Jing Qian and Jonathan Root and Venkatesh Saligrama},
  journal= {arXiv preprint arXiv:1502.01783},
  year   = {2015}
}

Comments

arXiv admin note: text overlap with arXiv:1405.0530

R2 v1 2026-06-22T08:23:28.877Z