Get Started
Research questionHow can numerical reconstruction-based anomaly detectors explain which feature ranges cause each alert?Reconstruction-based detectors can flag numerical outliers accurately, but their internal representations often do not show which feature ranges drove an alert. Practitioners therefore need explanations that connect an anomaly score to concrete, auditable conditions in the input.
Machine Learning
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Differentiable Interval Bottlenecks for Interpretable Anomaly Detection in Numerical DataThe source studies an autoencoder with differentiable, axis-aligned interval memberships learned directly from numerical data without discretization or binarization. Its intervals provide inspectable feature-range structure and label-free importance scores for candidate constraints, while reconstruction error remains the anomaly score. Evidence comes from 48 ADBench benchmarks compared with 22 baselines under a common normalized protocol, including inlier-only and contaminated-data settings.research paper · Sep 3, 2026
Related questions
How can unsupervised tabular anomaly detectors provide faithful feature-level explanations without costly post-hoc queries?How can quality assessment detect and localize anomalies introduced by enhancement in real-world user-generated images?How can training-free visual anomaly detection catch local defects and global rule violations without category-specific modeling?How can urban trajectory systems screen collective anomalies cheaply yet provide source-verifiable event details on demand?
Home
Topics
Search
Library