arxiv: 2604.13252 · v1 · submitted 2026-04-14 · 💻 cs.LG · cs.AI

Recognition: unknown

Out of Context: Reliability in Multimodal Anomaly Detection Requires Contextual Inference

Juan Miguel Valverde, Kamal Nasrollahi, Kevin Wilkinghoff, Neelu Madan, Radu Tudor Ionescu, Rafal Wisniewski, Thomas B. Moeslund, Wenwu Wang, Zheng-Hua Tan

Authors on Pith no claims yet

Pith reviewed 2026-05-10 16:06 UTC · model grok-4.3

classification 💻 cs.LG cs.AI

keywords anomaly detectionmultimodalcontextual inferencecross-modalreliabilityoperating conditionsstructural ambiguity

0 comments

The pith

Multimodal anomaly detection must separate context from observations to define abnormalities conditionally rather than against a single reference.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper argues that standard approaches to multimodal anomaly detection are unreliable because they model normality unconditionally, ignoring that what counts as anomalous often depends on the operating context. Existing methods treat all modalities the same way without distinguishing which provide context and which provide the signal to check. Reframing the task as cross-modal contextual inference, with asymmetric roles for modalities, allows abnormality to be assessed conditionally. A reader should care because real-world systems operate in dynamic environments where failing to account for context leads to false alarms or missed anomalies. This change would require new designs for models, evaluation, and benchmarks.

Core claim

The paper establishes that multimodal anomaly detection should be reframed as a cross-modal contextual inference problem. In this view, different modalities assume asymmetric roles: some capture the operating conditions as context, while others provide the observations whose normality is evaluated conditionally on that context. This replaces the assumption of a single global reference model of normality with conditional definitions of abnormality, addressing the structural ambiguity that arises when context and anomaly signals are mixed.

What carries the argument

The central mechanism is the asymmetric separation of modalities into context-inferring and observation-assessing streams to enable conditional abnormality detection.

If this is right

Models must be designed to infer context from one set of modalities and condition anomaly scores on it.
Evaluation protocols should test performance across varying contexts rather than averaged.
Benchmark datasets need to incorporate explicit context labels or variations.
This approach reduces instability in heterogeneous deployment environments.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

This reframing may apply to other multimodal tasks like classification or generation where context matters.
A testable extension would be to create datasets with labeled contexts and compare conditional vs unconditional detectors.
It connects to ideas in causal machine learning by treating context as a conditioning variable.

Load-bearing premise

The assumption that anomalies are frequently context-dependent, such that failing to separate context from observations creates unavoidable ambiguity in defining normality.

What would settle it

A dataset where context-dependent anomalies are present, and a symmetric multimodal method achieves comparable or better performance than a method that explicitly separates and conditions on context, would challenge the need for this reframing.

Figures

Figures reproduced from arXiv: 2604.13252 by Juan Miguel Valverde, Kamal Nasrollahi, Kevin Wilkinghoff, Neelu Madan, Radu Tudor Ionescu, Rafal Wisniewski, Thomas B. Moeslund, Wenwu Wang, Zheng-Hua Tan.

**Figure 1.** Figure 1: Context-dependent versus marginal abnormality. The red dotted line denotes an observation that is typical under context c1 but lies in the low-density tail of p(x | c2). When contextual information is ignored and a symmetric marginal reference p(x) = R p(x | c)p(c) dc is used, the same observation appears normal. Marginal modeling therefore entangles contextual variability with abnormal behavior, masking c… view at source ↗

**Figure 2.** Figure 2: Context-dependent interpretation of identical observations. A single visual input (person in a museum gallery) can yield different anomaly assessments depending on contextual variables with distinct observability and dependency structure. Environmental context (e.g., time) determines the operational state (open vs. closed) via policy (derived context). Identity constitutes modal context and is inferred fro… view at source ↗

read the original abstract

Anomaly detection aims to identify observations that deviate from expected behavior. Because anomalous events are inherently sparse, most frameworks are trained exclusively on normal data to learn a single reference model of normality. This implicitly assumes that normal behavior can be captured by a single, unconditional reference distribution. In practice, however, anomalies are often context-dependent: A specific observation may be normal under one operating condition, yet anomalous under another. As machine learning systems are deployed in dynamic and heterogeneous environments, these fixed-context assumptions introduce structural ambiguity, i.e., the inability to distinguish contextual variation from genuine abnormality under marginal modeling, leading to unstable performance and unreliable anomaly assessments. While modern sensing systems frequently collect multimodal data capturing complementary aspects of both system behavior and operating conditions, existing methods treat all data streams equally, without distinguishing contextual information from anomaly-relevant signals. As a result, abnormality is often evaluated without explicitly conditioning on operating conditions. We argue that multimodal anomaly detection should be reframed as a cross-modal contextual inference problem, in which modalities play asymmetric roles, separating context from observation, to define abnormality conditionally rather than relative to a single global reference. This perspective has implications for model design, evaluation protocols, and benchmark construction, and outline open research challenges toward robust, context-aware multimodal anomaly detection.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This is a perspective piece arguing that multimodal anomaly detection needs asymmetric context modeling to avoid mixing operating conditions with true anomalies, but it offers no examples or evidence to show the problem is real or fixable.

read the letter

The core pitch is that anomaly detection should stop treating all modalities as equal and instead use some as context to condition what counts as normal in the others. That split makes sense on paper for systems running in changing conditions, where a reading that looks off might just reflect a different operating state rather than a fault. The authors lay out how single-reference models create ambiguity here, and the logic tracks without obvious holes. They also flag downstream effects on benchmarks and evaluation, which is a useful reminder for anyone building these systems. What is missing is any concrete illustration. The abstract and argument stay general; there are no toy cases, dataset snippets, or even a sketch of how an asymmetric model would be trained or tested. Without that, the claim of structural ambiguity stays at the level of a reasonable hypothesis rather than a demonstrated gap. The paper does not contradict itself or hide assumptions, but it also does not move past restating the issue. This is aimed at people already working on multimodal anomaly detection in robotics, monitoring, or similar applied areas who are open to rethinking problem setup. A reader hunting for new methods or results will come away empty. It is worth sending to referees as a position paper so the community can discuss whether the reframing actually changes practice or just restates known context-sensitivity in fancier terms.

Referee Report

0 major / 1 minor

Summary. The manuscript is a perspective piece arguing that multimodal anomaly detection suffers from structural ambiguity under marginal modeling because existing methods treat all data streams equally without distinguishing contextual information from anomaly-relevant signals. It proposes reframing the task as a cross-modal contextual inference problem in which modalities play asymmetric roles, separating context from observation to define abnormality conditionally rather than relative to a single global reference, with implications for model design, evaluation, and benchmarks.

Significance. If pursued, the reframing could improve reliability of anomaly detection in dynamic, heterogeneous environments by promoting context-aware conditional modeling. The perspective identifies open research challenges that may usefully guide subsequent work on multimodal systems, though its impact will depend on whether the conceptual distinction leads to concrete methodological advances.

minor comments (1)

[Abstract] Abstract: the description of structural ambiguity is clear at a high level but would benefit from one brief, concrete example (e.g., a sensor reading that is normal under one operating condition but anomalous under another) to make the practical consequence more immediate for readers.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive assessment and the recommendation to accept the manuscript. The summary accurately reflects the central thesis of our perspective piece.

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The paper is a perspective piece advancing a conceptual reframing of multimodal anomaly detection as cross-modal contextual inference with asymmetric modality roles. Its central argument rests on stated premises about context-dependent anomalies and the limitations of marginal modeling in existing methods, without any equations, derivations, fitted parameters, or technical results that could reduce to self-referential inputs. No self-citations appear in the provided text, and the reasoning chain is self-contained as a call to reframe the problem rather than a derivation that collapses by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central argument rests on domain assumptions about how current anomaly detection systems operate and the prevalence of context dependence; no free parameters or new entities are introduced.

axioms (1)

domain assumption Anomalies are often context-dependent and current multimodal methods do not separate context from anomaly signals
Explicitly stated in the abstract as the source of structural ambiguity

pith-pipeline@v0.9.0 · 5560 in / 1162 out tokens · 91217 ms · 2026-05-10T16:06:21.017570+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

76 extracted references · 2 canonical work pages · 2 internal anchors

[1]

UBnormal: New Benchmark for Supervised Open-Set Video Anomaly Detection

Andra Acsintoae et al. “UBnormal: New Benchmark for Supervised Open-Set Video Anomaly Detection”. In:Proc. CVPR. 2022

2022
[2]

Aggarwal.Outlier Analysis

Charu C. Aggarwal.Outlier Analysis. 2nd. Springer, 2017

2017
[3]

Unsupervised real-time anomaly detection for streaming data

Subutai Ahmad et al. “Unsupervised real-time anomaly detection for streaming data”. In: Neurocomputing262 (2017)

2017
[4]

Flamingo: a Visual Language Model for Few-Shot Learning

Jean-Baptiste Alayrac et al. “Flamingo: a Visual Language Model for Few-Shot Learning”. In: Proc. NeurIPS. 2022

2022
[5]

IMAD-DS: A Dataset for Industrial Multi-Sensor Anomaly Detection Under Domain Shift Conditions

Davide Albertini et al. “IMAD-DS: A Dataset for Industrial Multi-Sensor Anomaly Detection Under Domain Shift Conditions”. In:Proc. DCASE. 2024

2024
[6]

USAD: UnSupervised Anomaly Detection on Multivariate Time Series

Julien Audibert et al. “USAD: UnSupervised Anomaly Detection on Multivariate Time Series”. In:Proc. KDD. 2020

2020
[7]

Multimodal Machine Learning: A Survey and Taxonomy

Tadas Baltrusaitis, Chaitanya Ahuja, and Louis-Philippe Morency. “Multimodal Machine Learning: A Survey and Taxonomy”. In:IEEE Trans. Pattern Anal. Mach. Intell.41.2 (2019)

2019
[8]

Towards Open Set Deep Networks

Abhijit Bendale and Terrance E. Boult. “Towards Open Set Deep Networks”. In:Proc. CVPR. 2016

2016
[9]

Representation Learning: A Review and New Perspectives

Yoshua Bengio, Aaron C. Courville, and Pascal Vincent. “Representation Learning: A Review and New Perspectives”. In:IEEE Trans. Pattern Anal. Mach. Intell.35.8 (2013)

2013
[10]

Beyond Dents and Scratches: Logical Constraints in Unsupervised Anomaly Detection and Localization

Paul Bergmann et al. “Beyond Dents and Scratches: Logical Constraints in Unsupervised Anomaly Detection and Localization”. In:Int. J. Comput. Vis.130.4 (2022)

2022
[11]

MVTec AD - A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection

Paul Bergmann et al. “MVTec AD - A Comprehensive Real-World Dataset for Unsupervised Anomaly Detection”. In:Proc. CVPR. 2019

2019
[12]

The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization

Paul Bergmann et al. “The MVTec 3D-AD Dataset for Unsupervised 3D Anomaly Detection and Localization”. In:Proc. VISIGRAPP. 2022

2022
[13]

Online Inference of Topics with Latent Dirichlet Allocation

Kevin Robert Canini, Lei Shi, and Thomas L. Griffiths. “Online Inference of Topics with Latent Dirichlet Allocation”. In:Proc. AISTATS. 2009

2009
[14]

Anomaly Detection under Distribution Shift

Tri Cao, Jiawen Zhu, and Guansong Pang. “Anomaly Detection under Distribution Shift”. In: Proc. ICCV. 2023

2023
[15]

Invariant Anomaly Detection under Distribution Shifts: A Causal Perspective

João B. S. Carvalho et al. “Invariant Anomaly Detection under Distribution Shifts: A Causal Perspective”. In:Proc. NeurIPS. 2023

2023
[16]

Anomaly detection: A survey

Varun Chandola, Arindam Banerjee, and Vipin Kumar. “Anomaly detection: A survey”. In: ACM Comput. Surv.41.3 (2009)

2009
[17]

Multimodal Industrial Anomaly Detection by Crossmodal Feature Mapping

Alex Costanzino et al. “Multimodal Industrial Anomaly Detection by Crossmodal Feature Mapping”. In:Proc. CVPR. 2024

2024
[18]

Description and Discussion on DCASE 2023 Challenge Task 2: First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring

Kota Dohi et al. “Description and Discussion on DCASE 2023 Challenge Task 2: First-Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring”. In:Proc. DCASE. 2023

2023
[19]

Description and Discussion on DCASE 2022 Challenge Task 2: Unsuper- vised Anomalous Sound Detection for Machine Condition Monitoring Applying Domain Generalization Techniques

Kota Dohi et al. “Description and Discussion on DCASE 2022 Challenge Task 2: Unsuper- vised Anomalous Sound Detection for Machine Condition Monitoring Applying Domain Generalization Techniques”. In:Proc. DCASE. 2022

2022
[20]

Continual learning for anomaly detection in surveillance videos

Keval Doshi and Yasin Yilmaz. “Continual learning for anomaly detection in surveillance videos”. In:Proc. CVPR Workshops. 2020

2020
[21]

Towards A Rigorous Science of Interpretable Machine Learning

Finale Doshi-Velez and Been Kim. “Towards a rigorous science of interpretable machine learning”. In:arXiv:1702.08608(2017)

work page internal anchor Pith review arXiv 2017
[22]

An introduction to ROC analysis

Tom Fawcett. “An introduction to ROC analysis”. In:Pattern Recognit. Lett.27.8 (2006)

2006
[23]

Deep multi-modal object detection and semantic segmentation for autonomous driving: Datasets, methods, and challenges

Di Feng et al. “Deep multi-modal object detection and semantic segmentation for autonomous driving: Datasets, methods, and challenges”. In:IEEE Trans. Intell. Transp. Syst.22.3 (2020)

2020
[24]

Recent Advances in Open Set Recognition: A Survey

Chuanxing Geng, Sheng-Jun Huang, and Songcan Chen. “Recent Advances in Open Set Recognition: A Survey”. In:IEEE Trans. Pattern Anal. Mach. Intell.43.10 (2021)

2021
[25]

Anomaly Detection in Video via Self-Supervised and Multi-Task Learning

Mariana-Iuliana Georgescu et al. “Anomaly Detection in Video via Self-Supervised and Multi-Task Learning”. In:Proc. CVPR. 2021

2021
[26]

A Baseline for Detecting Misclassified and Out-of- Distribution Examples in Neural Networks

Dan Hendrycks and Kevin Gimpel. “A Baseline for Detecting Misclassified and Out-of- Distribution Examples in Neural Networks”. In:Proc. ICLR. 2017. 10

2017
[27]

beta-V AE: Learning Basic Visual Concepts with a Constrained Variational Framework

Irina Higgins et al. “beta-V AE: Learning Basic Visual Concepts with a Constrained Variational Framework”. In:Proc. ICLR. 2017

2017
[28]

Adaptive Mixtures of Local Experts

Robert A. Jacobs et al. “Adaptive Mixtures of Local Experts”. In:Neural Comput.3.1 (1991)

1991
[29]

MMAD: A Comprehensive Benchmark for Multimodal Large Language Models in Industrial Anomaly Detection

Xi Jiang et al. “MMAD: A Comprehensive Benchmark for Multimodal Large Language Models in Industrial Anomaly Detection”. In:Proc. ICLR. 2025

2025
[30]

Hierarchical Mixtures of Experts and the EM Algorithm

Michael I. Jordan and Robert A. Jacobs. “Hierarchical Mixtures of Experts and the EM Algorithm”. In:Neural Comput.6.2 (1994)

1994
[31]

Description and Discussion on DCASE 2021 Challenge Task 2: Unsupervised Anomalous Detection for Machine Condition Monitoring Under Domain Shifted Conditions

Yohei Kawaguchi et al. “Description and Discussion on DCASE 2021 Challenge Task 2: Unsupervised Anomalous Detection for Machine Condition Monitoring Under Domain Shifted Conditions”. In:Proc. DCASE. 2021

2021
[32]

Auto-Encoding Variational Bayes

Diederik P. Kingma and Max Welling. “Auto-Encoding Variational Bayes”. In:Proc. ICLR. 2014

2014
[33]

Description and Discussion on DCASE2020 Challenge Task2: Unsu- pervised Anomalous Sound Detection for Machine Condition Monitoring

Yuma Koizumi et al. “Description and Discussion on DCASE2020 Challenge Task2: Unsu- pervised Anomalous Sound Detection for Machine Condition Monitoring”. In:Proc. DCASE. 2020

2020
[34]

A Review of Domain Adaptation without Target Labels

Wouter M. Kouw and Marco Loog. “A Review of Domain Adaptation without Target Labels”. In:IEEE Trans. Pattern Anal. Mach. Intell.43.3 (2021)

2021
[35]

Review of multimodal machine learning approaches in healthcare

Felix Krones et al. “Review of multimodal machine learning approaches in healthcare”. In:Inf. Fusion114 (2025)

2025
[36]

A Continual Learning Survey: Defying Forgetting in Classification Tasks

Matthias De Lange et al. “A Continual Learning Survey: Defying Forgetting in Classification Tasks”. In:IEEE Trans. Pattern Anal. Mach. Intell.44.7 (2022)

2022
[37]

Future Frame Prediction for Anomaly Detection - A New Baseline

Wen Liu et al. “Future Frame Prediction for Anomaly Detection - A New Baseline”. In:Proc. CVPR. 2018

2018
[38]

Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations

Francesco Locatello et al. “Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations”. In:Proc. ICML. 2019

2019
[39]

Self-supervised masked convolutional transformer block for anomaly detection

Neelu Madan et al. “Self-supervised masked convolutional transformer block for anomaly detection”. In:IEEE Trans. Pattern Anal. Mach. Intell.46.1 (2023)

2023
[40]

Temporal cues from socially unacceptable trajectories for anomaly detection

Neelu Madan et al. “Temporal cues from socially unacceptable trajectories for anomaly detection”. In:Proc. ICCV. 2021

2021
[41]

Analyzing a portion of the ROC curve

Donna Katzman McClish. “Analyzing a portion of the ROC curve”. In:Med. Decis. Mak.9.3 (1989)

1989
[42]

A unifying view on dataset shift in classification

Jose G. Moreno-Torres et al. “A unifying view on dataset shift in classification”. In:Pattern Recognit.45.1 (2012)

2012
[43]

Towards Understanding the Role of Over-Parametrization in Gener- alization of Neural Networks

Behnam Neyshabur et al. “Towards Understanding the Role of Over-Parametrization in Gener- alization of Neural Networks”. In:Proc. ICLR. 2020

2020
[44]

Multimodal Deep Learning

Jiquan Ngiam et al. “Multimodal Deep Learning”. In:Proc. ICML. 2011

2011
[45]

Description and Discussion on DCASE 2024 Challenge Task 2: First- Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring

Tomoya Nishida et al. “Description and Discussion on DCASE 2024 Challenge Task 2: First- Shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring”. In:Proc. DCASE. 2024

2024
[46]

Description and Discussion on DCASE 2025 Challenge Task 2: First- shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring

Tomoya Nishida et al. “Description and Discussion on DCASE 2025 Challenge Task 2: First- shot Unsupervised Anomalous Sound Detection for Machine Condition Monitoring”. In:Proc. DCASE. 2025

2025
[47]

Deep Learning for Anomaly Detection: A Review

Guansong Pang et al. “Deep Learning for Anomaly Detection: A Review”. In:ACM Comput. Surv.54.2 (2022)

2022
[48]

Normalizing Flows for Probabilistic Modeling and Inference

George Papamakarios et al. “Normalizing Flows for Probabilistic Modeling and Inference”. In: J. Mach. Learn. Res.22 (2021)

2021
[49]

A Multimodal Anomaly Detector for Robot-Assisted Feeding Using an LSTM-Based Variational Autoencoder

Daehyung Park, Yuuna Hoshi, and Charles C. Kemp. “A Multimodal Anomaly Detector for Robot-Assisted Feeding Using an LSTM-Based Variational Autoencoder”. In:IEEE Robotics Autom. Lett.3.2 (2018)

2018
[50]

Multimodal execution monitoring for anomaly detection during robot manipulation

Daehyung Park et al. “Multimodal execution monitoring for anomaly detection during robot manipulation”. In:Proc. ICRA. 2016

2016
[51]

Learning Transferable Visual Models From Natural Language Supervi- sion

Alec Radford et al. “Learning Transferable Visual Models From Natural Language Supervi- sion”. In:Proc. ICML. 2021

2021
[52]

Conditional Generative Models are Sufficient to Sample from Any Causal Effect Estimand

Md Musfiqur Rahman, Matt Jordan, and Murat Kocaoglu. “Conditional Generative Models are Sufficient to Sample from Any Causal Effect Estimand”. In:Proc. NeurIPS. 2024. 11

2024
[53]

Deep Multimodal Learning: A Survey on Recent Advances and Trends

Dhanesh Ramachandram and Graham W. Taylor. “Deep Multimodal Learning: A Survey on Recent Advances and Trends”. In:IEEE Signal Process. Mag.34.6 (2017)

2017
[54]

Self-supervised predictive convolutional attentive block for anomaly detection

Nicolae-C˘at˘alin Ristea et al. “Self-supervised predictive convolutional attentive block for anomaly detection”. In:Proc. CVPR. 2022

2022
[55]

Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead

Cynthia Rudin. “Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead”. In:Nat. Mach. Intell.1.5 (2019)

2019
[56]

A Unifying Review of Deep and Shallow Anomaly Detection

Lukas Ruff et al. “A Unifying Review of Deep and Shallow Anomaly Detection”. In:Proc. IEEE109.5 (2021)

2021
[57]

A Unified Survey on Anomaly, Novelty, Open-Set, and Out of-Distribution Detection: Solutions and Future Challenges

Mohammadreza Salehi et al. “A Unified Survey on Anomaly, Novelty, Open-Set, and Out of-Distribution Detection: Solutions and Future Challenges”. In:Trans. Mach. Learn. Res. (2022)

2022
[58]

Toward Open Set Recognition

Walter J. Scheirer et al. “Toward Open Set Recognition”. In:IEEE Trans. Pattern Anal. Mach. Intell.35.7 (2013)

2013
[59]

Toward Causal Representation Learning

Bernhard Schölkopf et al. “Toward Causal Representation Learning”. In:Proc. IEEE109.5 (2021)

2021
[60]

An overview of heart rate variability metrics and norms

Fred Shaffer and Jay P Ginsberg. “An overview of heart rate variability metrics and norms”. In:Front. Public Health5 (2017)

2017
[61]

Neural graphical models

Harsh Shrivastava and Urszula Chajewska. “Neural graphical models”. In:Proc. ECSQARU. 2023

2023
[62]

Eval: Explainable video anomaly localization

Ashish Singh, Michael J Jones, and Erik G Learned-Miller. “Eval: Explainable video anomaly localization”. In:Proc. CVPR. 2023

2023
[63]

Interpretable, Multidimensional, Multimodal Anomaly Detection with Negative Sampling for Detection of Device Failure

John Sipple. “Interpretable, Multidimensional, Multimodal Anomaly Detection with Negative Sampling for Detection of Device Failure”. In:Proc. ICML. V ol. 119. 2020

2020
[64]

Can you trust your model’s uncertainty? Evaluating predictive uncertainty under dataset shift

Jasper Snoek et al. “Can you trust your model’s uncertainty? Evaluating predictive uncertainty under dataset shift”. In:Proc. NeurIPS. 2019

2019
[65]

Learning Structured Output Representation using Deep Conditional Generative Models

Kihyuk Sohn, Honglak Lee, and Xinchen Yan. “Learning Structured Output Representation using Deep Conditional Generative Models”. In:Proc. NeurIPS. 2015

2015
[66]

Adaptive computation and machine learn- ing

Masashi Sugiyama and Motoaki Kawanabe.Machine Learning in Non-Stationary Environ- ments - Introduction to Covariate Shift Adaptation. Adaptive computation and machine learn- ing. 2012

2012
[67]

Real-world anomaly detection in surveillance videos

Waqas Sultani, Chen Chen, and Mubarak Shah. “Real-world anomaly detection in surveillance videos”. In:Proc. CVPR. 2018

2018
[68]

CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances

Jihoon Tack et al. “CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances”. In:Proc. NeurIPS. 2020

2020
[69]

On Disentangled Representations Learned from Correlated Data

Frederik Träuble et al. “On Disentangled Representations Learned from Correlated Data”. In: Proc. ICML. V ol. 139. 2021

2021
[70]

Multimodal Industrial Anomaly Detection via Hybrid Fusion

Yue Wang et al. “Multimodal Industrial Anomaly Detection via Hybrid Fusion”. In:Proc. CVPR. 2023

2023
[71]

How Much Does Machine Identity Matter in Anomalous Sound Detection at Test Time?

Kevin Wilkinghoff, Keisuke Imoto, and Zheng-Hua Tan. “How Much Does Machine Identity Matter in Anomalous Sound Detection at Test Time?” In:arXiv:2602.16253(2026)

work page internal anchor Pith review arXiv 2026
[72]

Handling Domain Shifts for Anomalous Sound Detection: A Review of DCASE-Related Work

Kevin Wilkinghoff et al. “Handling Domain Shifts for Anomalous Sound Detection: A Review of DCASE-Related Work”. In:Proc. DCASE. 2025

2025
[73]

Local Density-Based Anomaly Score Normalization for Domain Generalization

Kevin Wilkinghoff et al. “Local Density-Based Anomaly Score Normalization for Domain Generalization”. In:IEEE Trans. Audio, Speech, Lang. Process.33 (2025)

2025
[74]

Can We Evaluate Domain Adaptation Models Without Target-Domain Labels?

Jianfei Yang et al. “Can We Evaluate Domain Adaptation Models Without Target-Domain Labels?” In:Proc. ICLR. 2024

2024
[75]

Deep learning and its applications to machine health monitoring

Rui Zhao et al. “Deep learning and its applications to machine health monitoring”. In:Mech. Syst. Signal Process.115 (2019)

2019
[76]

Domain Generalization: A Survey

Kaiyang Zhou et al. “Domain Generalization: A Survey”. In:IEEE Trans. Pattern Anal. Mach. Intell.45.4 (2023). 12

2023