mlsanityData to 14 Oct 2026

Perfect density models cannot guarantee anomaly detection

arXiv 2012.03808 ↗First posted 7 Dec 2020cs.LG stat.MLcited by 39

Not on the map. Why?

Abstract (arXiv)

Thanks to the tractability of their likelihood, several deep generative models show promise for seemingly straightforward but important applications like anomaly detection, uncertainty estimation, and active learning. However, the likelihood values empirically attributed to anomalies conflict with the expectations these proposed applications suggest. In this paper, we take a closer look at the behavior of distribution densities through the lens of reparametrization and show that these quantities carry less meaningful information than previously thought, beyond estimation issues or the curse of dimensionality. We conclude that the use of these likelihoods for anomaly detection relies on strong and implicit hypotheses, and highlight the necessity of explicitly formulating these assumptions for reliable anomaly detection.

Key facts. Cited by 39 papers in the map (data through 14 Oct 2026). First posted on arXiv 7 Dec 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 10; 2022: 8; 2023: 6; 2024: 9; 2025: 3; 2026: 3.

39 papers.

Sort:

By area:

  1. 2022Asymmetric Student-Teacher Networks for Industrial Anomaly Detectioncited by 117
  2. 2021Understanding Failures in Out-of-Distribution Detection with Deep Generative Modelscited by 89
  3. 2021Graph Posterior Network: Bayesian Predictive Uncertainty for Node Classificationcited by 80
  4. 2022Diagnosing and Fixing Manifold Overfitting in Deep Generative Modelscited by 28
  5. 2021Autoencoding Under Normalization Constraintscited by 26
  6. 2023Feature Likelihood Divergence: Evaluating the Generalization of Generative Models Using Samplescited by 25
  7. 2024A Geometric Explanation of the Likelihood OOD Detection Paradoxcited by 22
  8. 2021Do We Really Need to Learn Representations from In-domain Data for Outlier Detection?cited by 16
  9. 2021Entropic Issues in Likelihood-Based OOD Detectioncited by 16
  10. 2021Natural Posterior Network: Deep Bayesian Uncertainty for Exponential Family Distributionscited by 13
  11. 2025A Survey on Diffusion Models for Anomaly Detectioncited by 13
  12. 2024What makes an image realistic?cited by 10
27 more
  1. 2023Anomaly Detection with Variance Stabilized Density Estimationcited by 8
  2. 2024PQMass: Probabilistic Assessment of the Quality of Generative Models using Probability Mass Estimationcited by 8
  3. 2021Innovations Autoencoder and its Application in One-class Anomalous Sequence Detectioncited by 7
  4. 2021Understanding Out-of-distribution: A Perspective of Data Dynamicscited by 7
  5. 2023ODIM: Outlier Detection via Likelihood of Under-Fitted Generative Modelscited by 7
  6. 2021New Methods and Datasets for Group Anomaly Detection From Fundamental Physicscited by 6
  7. 2022Inference and Sampling for Archimax Copulascited by 5
  8. 2022Invariant Representation Driven Neural Classifier for Anti-QCD Jet Taggingcited by 4
  9. 2025Generative Artificial Intelligence for Internet of Things Computing: A Systematic Surveycited by 3
  10. 2024Approximations to the Fisher Information Metric of Deep Generative Models for Out-Of-Distribution Detectioncited by 2
  11. 2022Repairing Systematic Outliers by Learning Clean Subspaces in VAEscited by 1
  12. 2022Evaluating Out-of-Distribution Detectors Through Adversarial Generation of Outlierscited by 1
  13. 2023Master's Thesis: Out-of-distribution Detection with Energy-based Modelscited by 1
  14. 2023Vacant Holes for Unsupervised Detection of the Outliers in Compact Latent Representationcited by 1
  15. 2024ALTBI: Constructing Improved Outlier Detection Models via Optimization of Inlier-Memorization Effectcited by 1
  16. 2024Channel Matters: Estimating Channel Influence for Multivariate Time Seriescited by 1
  17. 2024Why CNN Features Are not Gaussian: A Statistical Anatomy of Deep Representationscited by 1
  18. 2025Time-reversed Flow Matching with Worst Transport in High-dimensional Latent Space for Image Anomaly Detectioncited by 1
  19. 2026MIRA: A Score for Conditional Distribution Accuracy and Model Comparisoncited by 1
  20. 2021DAAIN: Detection of Anomalous and Adversarial Input using Normalizing Flowscited by 0
  21. 2022Maximum Likelihood on the Joint (Data, Condition) Distribution for Solving Ill-Posed Problems with Conditional Flow Modelscited by 0
  22. 2022On the Connection of Generative Models and Discriminative Models for Anomaly Detectioncited by 0
  23. 2023Deep machine learning for meteor monitoring: advances with transfer learning and gradient-weighted class activation mappingcited by 0
  24. 2024Rethinking Test-time Likelihood: The Likelihood Path Principle and Its Application to OOD Detectioncited by 0
  25. 2024Grid Monitoring with Synchro-Waveform and AI Foundation Model Technologiescited by 0
  26. 2026Mitigating the Likelihood Paradox in Flow-based OOD Detection via Entropy Manipulationcited by 0
  27. 2026Why the Counterintuitive Phenomenon of Likelihood Rarely Appears in Tabular Anomaly Detection with Deep Generative Models?cited by 0