Robust, Deep and Inductive Anomaly Detection

Raghavendra Chalapathy, Aditya Krishna Menon, Sanjay Chawla

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Citations (Scopus)

Abstract

PCA is a classical statistical technique whose simplicity and maturity has seen it find widespread use for anomaly detection. However, it is limited in this regard by being sensitive to gross perturbations of the input, and by seeking a linear subspace that captures normal behaviour. The first issue has been dealt with by robust PCA, a variant of PCA that explicitly allows for some data points to be arbitrarily corrupted; however, this does not resolve the second issue, and indeed introduces the new issue that one can no longer inductively find anomalies on a test set. This paper addresses both issues in a single model, the robust autoencoder. This method learns a nonlinear subspace that captures the majority of data points, while allowing for some data to have arbitrary corruption. The model is simple to train and leverages recent advances in the optimisation of deep neural networks. Experiments on a range of real-world datasets highlight the model’s effectiveness.

Original languageEnglish
Title of host publicationMachine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2017, Proceedings
PublisherSpringer Verlag
Pages36-51
Number of pages16
ISBN (Print)9783319712482
DOIs
Publication statusPublished - 1 Jan 2017
EventEuropean Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2017 - Skopje, Macedonia, The Former Yugoslav Republic of
Duration: 18 Sep 201722 Sep 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10534 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

OtherEuropean Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2017
CountryMacedonia, The Former Yugoslav Republic of
CitySkopje
Period18/9/1722/9/17

Fingerprint

Anomaly Detection
Subspace
Test Set
Gross
Leverage
Anomaly
Resolve
Simplicity
Model
Neural Networks
Perturbation
Optimization
Arbitrary
Range of data
Experiment
Experiments

Keywords

  • Anomaly detection
  • Autoencoders
  • Deep learning
  • Outlier detection
  • Robust PCA

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Chalapathy, R., Menon, A. K., & Chawla, S. (2017). Robust, Deep and Inductive Anomaly Detection. In Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2017, Proceedings (pp. 36-51). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10534 LNAI). Springer Verlag. https://doi.org/10.1007/978-3-319-71249-9_3

Robust, Deep and Inductive Anomaly Detection. / Chalapathy, Raghavendra; Menon, Aditya Krishna; Chawla, Sanjay.

Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2017, Proceedings. Springer Verlag, 2017. p. 36-51 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10534 LNAI).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Chalapathy, R, Menon, AK & Chawla, S 2017, Robust, Deep and Inductive Anomaly Detection. in Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2017, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10534 LNAI, Springer Verlag, pp. 36-51, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML PKDD 2017, Skopje, Macedonia, The Former Yugoslav Republic of, 18/9/17. https://doi.org/10.1007/978-3-319-71249-9_3
Chalapathy R, Menon AK, Chawla S. Robust, Deep and Inductive Anomaly Detection. In Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2017, Proceedings. Springer Verlag. 2017. p. 36-51. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-71249-9_3
Chalapathy, Raghavendra ; Menon, Aditya Krishna ; Chawla, Sanjay. / Robust, Deep and Inductive Anomaly Detection. Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2017, Proceedings. Springer Verlag, 2017. pp. 36-51 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{0572280cb71a4d4e8d8bf15d7d7f18f0,
title = "Robust, Deep and Inductive Anomaly Detection",
abstract = "PCA is a classical statistical technique whose simplicity and maturity has seen it find widespread use for anomaly detection. However, it is limited in this regard by being sensitive to gross perturbations of the input, and by seeking a linear subspace that captures normal behaviour. The first issue has been dealt with by robust PCA, a variant of PCA that explicitly allows for some data points to be arbitrarily corrupted; however, this does not resolve the second issue, and indeed introduces the new issue that one can no longer inductively find anomalies on a test set. This paper addresses both issues in a single model, the robust autoencoder. This method learns a nonlinear subspace that captures the majority of data points, while allowing for some data to have arbitrary corruption. The model is simple to train and leverages recent advances in the optimisation of deep neural networks. Experiments on a range of real-world datasets highlight the model’s effectiveness.",
keywords = "Anomaly detection, Autoencoders, Deep learning, Outlier detection, Robust PCA",
author = "Raghavendra Chalapathy and Menon, {Aditya Krishna} and Sanjay Chawla",
year = "2017",
month = "1",
day = "1",
doi = "10.1007/978-3-319-71249-9_3",
language = "English",
isbn = "9783319712482",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "36--51",
booktitle = "Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2017, Proceedings",

}

TY - GEN

T1 - Robust, Deep and Inductive Anomaly Detection

AU - Chalapathy, Raghavendra

AU - Menon, Aditya Krishna

AU - Chawla, Sanjay

PY - 2017/1/1

Y1 - 2017/1/1

N2 - PCA is a classical statistical technique whose simplicity and maturity has seen it find widespread use for anomaly detection. However, it is limited in this regard by being sensitive to gross perturbations of the input, and by seeking a linear subspace that captures normal behaviour. The first issue has been dealt with by robust PCA, a variant of PCA that explicitly allows for some data points to be arbitrarily corrupted; however, this does not resolve the second issue, and indeed introduces the new issue that one can no longer inductively find anomalies on a test set. This paper addresses both issues in a single model, the robust autoencoder. This method learns a nonlinear subspace that captures the majority of data points, while allowing for some data to have arbitrary corruption. The model is simple to train and leverages recent advances in the optimisation of deep neural networks. Experiments on a range of real-world datasets highlight the model’s effectiveness.

AB - PCA is a classical statistical technique whose simplicity and maturity has seen it find widespread use for anomaly detection. However, it is limited in this regard by being sensitive to gross perturbations of the input, and by seeking a linear subspace that captures normal behaviour. The first issue has been dealt with by robust PCA, a variant of PCA that explicitly allows for some data points to be arbitrarily corrupted; however, this does not resolve the second issue, and indeed introduces the new issue that one can no longer inductively find anomalies on a test set. This paper addresses both issues in a single model, the robust autoencoder. This method learns a nonlinear subspace that captures the majority of data points, while allowing for some data to have arbitrary corruption. The model is simple to train and leverages recent advances in the optimisation of deep neural networks. Experiments on a range of real-world datasets highlight the model’s effectiveness.

KW - Anomaly detection

KW - Autoencoders

KW - Deep learning

KW - Outlier detection

KW - Robust PCA

UR - http://www.scopus.com/inward/record.url?scp=85040249118&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85040249118&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-71249-9_3

DO - 10.1007/978-3-319-71249-9_3

M3 - Conference contribution

AN - SCOPUS:85040249118

SN - 9783319712482

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 36

EP - 51

BT - Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2017, Proceedings

PB - Springer Verlag

ER -