Deep Semi-Supervised Anomaly Detection

By Lukas Ruff, Robert A. Vandermeulen, Nico Görnitz, Alexander Binder, Emmanuel Müller, K. Müller, Marius Kloft. International Conference on Learning Representations, 2019

704 citations recorded by the source corpus.

Deep approaches to anomaly detection have recently shown promising results over shallow methods on large and complex datasets. Typically anomaly detection is treated as an unsupervised learning problem. In practice however, one may have---in addition to a large set of unlabeled samples---access to a small pool of labeled samples, e.g. a subset verified by some domain expert as being normal or anomalous. Semi-supervised approaches to anomaly detection aim to utilize such labeled samples, but most proposed methods are limited to merely including labeled normal samples. Only a few methods take advantage of labeled anomalies, with existing deep approaches being domain-specific. In this work we present Deep SAD, an end-to-end deep methodology for general semi-supervised anomaly detection. We further introduce an information-theoretic framework for deep anomaly detection based on the idea that the entropy of the latent distribution for normal data should be lower than the entropy of the anomalous distribution, which can serve as a theoretical interpretation for our method. In extensive experiments on MNIST, Fashion-MNIST, and CIFAR-10, along with other anomaly detection benchmark datasets, we demonstrate that our method is on par or outperforms shallow, hybrid, and deep competitors, yielding appreciable performance improvements even when provided with only little labeled data.

Subfields: Artificial Intelligence, Computer Networks and Communications, Epidemiology

View Paper: https://www.semanticscholar.org/paper/4c1abd8969fc1c360f50373f6552bcfb3cc408b7

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

In Computer Networks and Communications

In Epidemiology

Citations in the corpus

A Unifying Review of Deep and Shallow Anomaly Detection (2020, methodology): “However, including known anomalous examples, if available, can significantly improve the detection performance of a model [144], [224], [228]–[231].”
Multiresolution Knowledge Distillation for Anomaly Detection (2020, methodology): “Due to the unsupervised nature of the AD problem and the restricted data access, availability of just the normal data in training, the majority of methods [36, 31, 40, 18, 34] model the normal data abstraction by extracting semantically meaningful latent features.”
ADBench: Anomaly Detection Benchmark (2022, methodology): “Following previous works [151, 152], we set one of the multi-classes as normal, downsample the remaining classes to 5% of the total instances as anomalies, and report the average results over all the respective classes.”
End-to-End Reconstruction-Classification Learning for Face Forgery Detection (2022, neutral): “As previous studies [5, 36] indicate that real samples possess a relatively compact distribution, the”
CSI: Novelty Detection via Contrastive Learning on Distributionally Shifted Instances (2020, background): “There are other settings, e . g ., semi-supervised setting - the detector can access a small subset of out-of-distribution samples [25, 54], or supervised setting - the detector knows the target out-of-distribution, but we do not consider those settings in this paper.”
SSD: A Unified Framework for Self-Supervised Outlier Detection (2021, neutral): “2 Deep-SAD (Ruff et al., 2019) – – – – – – – – – – 77.”
Deep Learning for Anomaly Detection (2020, background): “[175] Mutant activity detection p53 Mutants [35] 16,772 5,408 0.48% Tabular [114] Internet ads detection AD [13] 3,279 1,555 14.00% Tabular [114] Disease detection Thyroid [13] 7,200 21 7.40% Tabular [115,116,134,179] Disease detection Arrhythmia [13] 452 279 14.60% Tabular [117,134,179] Defect dete”
Patch SVDD: Patch-level SVDD for Anomaly Detection and Segmentation (2020, methodology): “[5] re-interpreted Deep SVDD in an information-theoretic perspective and applied to semi-supervised scenarios.”
A Comprehensive Survey on Graph Anomaly Detection With Deep Learning (2021, background): “Amongst earlier works in this area, the detection methods reply heavily on handcrafted feature engineering or statistical models built by domain experts [24], [36]–[38].”
OpenGAN: Open-Set Recognition via Open Data Generation (2021, neutral): “[15, 25, 43] reformulate the problem with the concept of “outlier exposure” which allows methods to access some outlier data as open-training examples.”