A survey on unsupervised outlier detection in high‐dimensional numerical data

By Arthur Zimek, Erich Schubert, Hans-Peter Kriegel. Statistical analysis and data mining, 2012

858 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/b65fc8f5e7329f0476bc7280f0ef6b91a8c8484b

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 Statistics and Probability

Citations in the corpus

Density‐based clustering (2019, background): “However, with the same amount of data a density estimate that is not corrected for the dimensionality—like the distance to the k th nearest neighbor or the number of points within a ball of fixed radius r or within a bin (hypercube) of fixed side length, which is what most traditional methods are us”
Memorizing Normality to Detect Anomaly: Memory-Augmented Deep Autoencoder for Unsupervised Anomaly Detection (2019, methodology): “Unsupervised clustering methods, such as the k-means method and Gaussian Mixture Models (GMM) [47, 40], have also been applied to build a detailed profile of the normal data for identifying the anomalies.”
Deep Structured Energy Based Models for Anomaly Detection (2016, methodology): “A variety of methods can be found in the survey (Zimek et al., 2012).”
Learning to group discrete graphical patterns (2017, neutral): “The representation is extremely high-dimensional, which can lead to unreliable distances when comparing elements [Zimek et al. 2012].”
A Unifying Review of Deep and Shallow Anomaly Detection (2020, background): “The greater the noise, the harder it is to accurately estimate the ground-truth level sets of P since informative normal features get obfuscated [165].”
Deep Learning for Anomaly Detection (2020, background): “As a popular area, a number of studies [2, 4, 16, 28, 53, 62, 178] have been dedicated to the categorization and review of anomaly detection techniques.”
Robust Anomaly Detection for Multivariate Time Series through Stochastic Recurrent Neural Network (2019, background): “However, despite the rich literature in multivariate time series anomaly detection in different areas [4, 6, 11, 16, 17, 20, 26, 27], previous studies either take deterministic approaches [4, 6, 11] to model time series, or take stochastic approach but ignore the temporal dependence of observations ”
A high-bias, low-variance introduction to Machine Learning for physicists (2018, neutral): “This in turn leads to increased errors for pairwise similarity and distance measures and thus tends to “blur” distances between data points (Domingos, 2012; Kriegel et al., 2009; Zimek et al., 2012).”
Statistical Analysis of Nearest Neighbor Methods for Anomaly Detection (2019, background): “From another perspective, [29] discusses the various effects of dimensionality in the context of anomaly detection.”
Loda: Lightweight on-line detector of anomalies (2016, background): “We refer to Zimek et al. (2012) for a comprehensive review of the prior art.”
Deep Autoencoding Gaussian Mixture Model for Unsupervised Anomaly Detection (2018, methodology): “In the training phase with unknown mixture-component distribution φ, mixture means µ, and mixture covariance Σ, the estimation network estimates the parameters of GMM and evaluates the likelihood/energy for samples without alternating procedures such as EM (Zimek et al. (2012)).”
Classification-Reconstruction Learning for Open-Set Recognition (2018, background): “This bottlenecking is crucial, because outliers are harder to detect in higher dimensional feature spaces due to concentration on the sphere [52].”
Efficient GAN-Based Anomaly Detection (2018, methodology): “Popular techniques utilize clustering approaches or nearest neighbor methods (Xiong et al., 2011; Zimek et al., 2012) and one class classification approaches that learn a discriminative boundary around normal data, such as one-class SVMs (Yunqiang Chen & Huang, 2001).”
High-dimensional and large-scale anomaly detection using a linear one-class SVM with deep learning (2016, background): “detection due to the following factors [7]: (i) Exponential search”
DBSCAN Revisited, Revisited (2017, background): “Independent of the algorithm, the parameter ε of DBSCAN becomes hard to choose in highdimensional data due to the loss of contrast in distances [9, 25, 48].”
Robust Anomaly Detection and Backdoor Attack Detection Via Differential Privacy (2019, background): “Previous work mostly achieves outlier detection using unsupervised learning methods ([25, 26, 27]), while novelty detection typically assumes a normal dataset is available for training, and is realized by semi-supervised learning ([28, 29]).”
Outlier Analysis (2013, background): “It has been claimed in [514] as an apparently new insight, that the major reason for difficulty in high dimensional outlier analysis is not the concentration of distances, but the masking effects of the locally noisy and irrelevant nature of some of the dimensions, and that the literature has failed”