A Unifying Review of Deep and Shallow Anomaly Detection

By Lukas Ruff, Jacob R. Kauffmann, Robert A. Vandermeulen, Grgoire Montavon, Wojciech Samek, Marius Kloft, Thomas G. Dietterich, Klaus‐Robert Müller . Proceedings of the IEEE, 2020

1,067 citations recorded by the source corpus.

Deep learning approaches to anomaly detection (AD) have recently improved the state of the art in detection performance on complex data sets, such as large collections of images or text. These results have sparked a renewed interest in the AD problem and led to the introduction of a great variety of new methods. With the emergence of numerous such methods, including approaches based on generative models, one-class classification, and reconstruction, there is a growing need to bring methods of this field into a systematic and unified perspective. In this review, we aim to identify the common underlying principles and the assumptions that are often made implicitly by various methods. In particular, we draw connections between classic “shallow” and novel deep approaches and show how this relation might cross-fertilize or extend both directions. We further provide an empirical assessment of major existing methods that are enriched by the use of recent explainability techniques and present specific worked-through examples together with practical advice. Finally, we outline critical open challenges and identify specific paths for future research in AD.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/30b99ae0682d42a2010be401dd1d8f7baca9bb5f

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

ADBench: Anomaly Detection Benchmark (2022, background): “Many unsupervised methods have been proposed in the last few decades [3, 15, 129, 150, 198], which can be roughly categorized into shallow and deep (neural network) methods.”
Deep Isolation Forest for Anomaly Detection (2022, background): “Surveys and comparative studies can be found in [2], [16], [17].”
AnomalyCLIP: Object-agnostic Prompt Learning for Zero-shot Anomaly Detection (2023, neutral): “Existing AD approaches typically assume that training examples in a target application domain are available for learning the detection models (Pang et al., 2021b; Ruff et al., 2021).”
Transformers in Time Series: A Survey (2022, positive): “Deep learning also triggers novel developments for anomaly detection [Ruff et al., 2021].”
Generalized Out-of-Distribution Detection: A Survey (2021, neutral): “…Oursensory/semantic dichotomy for the AD sub-task deﬁnition comes from the low-level sensory anomalies and high-level semantic anomalies that are introduced in Ahmed and Courville (2020) and highlighted in the recent AD survey (Ruff et al., 2021), to reﬂect the rise of deep learning.”
CutPaste: Self-Supervised Learning for Anomaly Detection and Localization (2021, methodology): “We also compare with previous works, including deep one-class classifier (DOCC) [45], uninformed student [5], and patch SVDD [61].”
OpenOOD: Benchmarking Generalized Out-of-Distribution Detection (2022, background): “The AD task can also be divided into unsupervised AD, and (semi-)supervised AD in regard to the data supervision [37].”
SPot-the-Difference Self-Supervised Pre-training for Anomaly Detection and Segmentation (2022, methodology): “Many recent methods are proposed to detect low-level texture anomalies [35], such as scratches and cracks, which are common cases in industrial visual inspection [15,31,34,44].”
Explaining Deep Neural Networks and Beyond: A Review of Methods and Applications (2021, background): “However, for unsupervised problems, such as anomaly detection or clustering, although deep models have also reached successes [27], [155], [156], shallow models (e.”
DCdetector: Dual Attention Contrastive Representation Learning for Time Series Anomaly Detection (2023, methodology): “Unsupervised methods are popular without strict requirements on labeled data, including one class classification-based, probabilisticbased, distance-based, forecasting-based, reconstruction-based approaches [11, 24, 42, 59, 66, 85, 92].”
Unsupervised Deep Anomaly Detection for Multi-Sensor Time-Series Signals (2021, background): “Due to above challenges, there is a growing trend to use unsupervised learning approaches for anomaly detection compared with semi-supervised and supervised learning approaches since unsupervised methods can handle the imbalanced and unlabeled data in a more principled way [5], [6], [7], [8], [9].”