Unsupervised Surrogate Anomaly Detection
By: Simon Klüttermann, Tim Katzke, Emmanuel Müller
Potential Business Impact:
Finds weird things in data.
In this paper, we study unsupervised anomaly detection algorithms that learn a neural network representation, i.e. regular patterns of normal data, which anomalies are deviating from. Inspired by a similar concept in engineering, we refer to our methodology as surrogate anomaly detection. We formalize the concept of surrogate anomaly detection into a set of axioms required for optimal surrogate models and propose a new algorithm, named DEAN (Deep Ensemble ANomaly detection), designed to fulfill these criteria. We evaluate DEAN on 121 benchmark datasets, demonstrating its competitive performance against 19 existing methods, as well as the scalability and reliability of our method.
Similar Papers
Fairness-aware Anomaly Detection via Fair Projection
Machine Learning (CS)
Makes AI fairer by spotting bad data for everyone.
Bridging Unsupervised and Semi-Supervised Anomaly Detection: A Theoretically-Grounded and Practical Framework with Synthetic Anomalies
Machine Learning (Stat)
Finds hidden problems by creating fake ones.
A systematic literature review of unsupervised learning algorithms for anomalous traffic detection based on flows
Cryptography and Security
Finds hidden internet dangers without knowing them first.