Abstract

Learning expressive low-dimensional representations of ultrahigh-dimensional data, e.g., data with thousands/millions of features, has been a major way to enable learning methods to address the curse of dimensionality. However, existing unsupervised representation learning methods mainly focus on preserving the data regularity information and learning the representations independently of subsequent outlier detection methods, which can result in suboptimal and unstable performance of detecting irregularities (i.e., outliers). This paper introduces a ranking model-based framework, called RAMODO, to address this issue. RAMODO unifies representation learning and outlier detection to learn low-dimensional representations that are tailored for a state-of-the-art outlier detection approach - the random distance-based approach. This customized learning yields more optimal and stable representations for the targeted outlier detectors. Additionally, RAMODO can leverage little labeled data as prior knowledge to learn more expressive and application-relevant representations. We instantiate RAMODO to an efficient method called REPEN to demonstrate the performance of RAMODO. Extensive empirical results on eight real-world ultrahigh dimensional data sets show that REPEN (i) enables a random distance-based detector to obtain significantly better AUC performance and two orders of magnitude speedup; (ii) performs substantially better and more stably than four state-of-the-art representation learning methods; and (iii) leverages less than 1% labeled data to achieve up to 32% AUC improvement.

Original languageEnglish (US)
Title of host publicationKDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining
PublisherAssociation for Computing Machinery
Pages2041-2050
Number of pages10
ISBN (Print)9781450355520
DOIs
StatePublished - Jul 19 2018
Event24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2018 - London, United Kingdom
Duration: Aug 19 2018Aug 23 2018

Other

Other24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2018
CountryUnited Kingdom
CityLondon
Period8/19/188/23/18

Fingerprint

Detectors

Keywords

  • Dimension Reduction
  • Outlier Detection
  • Representation Learning
  • Ultrahigh-dimensional Data

ASJC Scopus subject areas

  • Software
  • Information Systems

Cite this

Pang, G., Chen, L., Cao, L., & Liu, H. (2018). Learning representations of ultrahigh-dimensional data for random distance-based outlier detection. In KDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (pp. 2041-2050). Association for Computing Machinery. https://doi.org/10.1145/3219819.3220042

Learning representations of ultrahigh-dimensional data for random distance-based outlier detection. / Pang, Guansong; Chen, Ling; Cao, Longbing; Liu, Huan.

KDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery, 2018. p. 2041-2050.

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

Pang, G, Chen, L, Cao, L & Liu, H 2018, Learning representations of ultrahigh-dimensional data for random distance-based outlier detection. in KDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery, pp. 2041-2050, 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, KDD 2018, London, United Kingdom, 8/19/18. https://doi.org/10.1145/3219819.3220042
Pang G, Chen L, Cao L, Liu H. Learning representations of ultrahigh-dimensional data for random distance-based outlier detection. In KDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery. 2018. p. 2041-2050 https://doi.org/10.1145/3219819.3220042
Pang, Guansong ; Chen, Ling ; Cao, Longbing ; Liu, Huan. / Learning representations of ultrahigh-dimensional data for random distance-based outlier detection. KDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Association for Computing Machinery, 2018. pp. 2041-2050
@inproceedings{0e1339c0e4924d93b1695af2f9e3e03c,
title = "Learning representations of ultrahigh-dimensional data for random distance-based outlier detection",
abstract = "Learning expressive low-dimensional representations of ultrahigh-dimensional data, e.g., data with thousands/millions of features, has been a major way to enable learning methods to address the curse of dimensionality. However, existing unsupervised representation learning methods mainly focus on preserving the data regularity information and learning the representations independently of subsequent outlier detection methods, which can result in suboptimal and unstable performance of detecting irregularities (i.e., outliers). This paper introduces a ranking model-based framework, called RAMODO, to address this issue. RAMODO unifies representation learning and outlier detection to learn low-dimensional representations that are tailored for a state-of-the-art outlier detection approach - the random distance-based approach. This customized learning yields more optimal and stable representations for the targeted outlier detectors. Additionally, RAMODO can leverage little labeled data as prior knowledge to learn more expressive and application-relevant representations. We instantiate RAMODO to an efficient method called REPEN to demonstrate the performance of RAMODO. Extensive empirical results on eight real-world ultrahigh dimensional data sets show that REPEN (i) enables a random distance-based detector to obtain significantly better AUC performance and two orders of magnitude speedup; (ii) performs substantially better and more stably than four state-of-the-art representation learning methods; and (iii) leverages less than 1{\%} labeled data to achieve up to 32{\%} AUC improvement.",
keywords = "Dimension Reduction, Outlier Detection, Representation Learning, Ultrahigh-dimensional Data",
author = "Guansong Pang and Ling Chen and Longbing Cao and Huan Liu",
year = "2018",
month = "7",
day = "19",
doi = "10.1145/3219819.3220042",
language = "English (US)",
isbn = "9781450355520",
pages = "2041--2050",
booktitle = "KDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining",
publisher = "Association for Computing Machinery",

}

TY - GEN

T1 - Learning representations of ultrahigh-dimensional data for random distance-based outlier detection

AU - Pang, Guansong

AU - Chen, Ling

AU - Cao, Longbing

AU - Liu, Huan

PY - 2018/7/19

Y1 - 2018/7/19

N2 - Learning expressive low-dimensional representations of ultrahigh-dimensional data, e.g., data with thousands/millions of features, has been a major way to enable learning methods to address the curse of dimensionality. However, existing unsupervised representation learning methods mainly focus on preserving the data regularity information and learning the representations independently of subsequent outlier detection methods, which can result in suboptimal and unstable performance of detecting irregularities (i.e., outliers). This paper introduces a ranking model-based framework, called RAMODO, to address this issue. RAMODO unifies representation learning and outlier detection to learn low-dimensional representations that are tailored for a state-of-the-art outlier detection approach - the random distance-based approach. This customized learning yields more optimal and stable representations for the targeted outlier detectors. Additionally, RAMODO can leverage little labeled data as prior knowledge to learn more expressive and application-relevant representations. We instantiate RAMODO to an efficient method called REPEN to demonstrate the performance of RAMODO. Extensive empirical results on eight real-world ultrahigh dimensional data sets show that REPEN (i) enables a random distance-based detector to obtain significantly better AUC performance and two orders of magnitude speedup; (ii) performs substantially better and more stably than four state-of-the-art representation learning methods; and (iii) leverages less than 1% labeled data to achieve up to 32% AUC improvement.

AB - Learning expressive low-dimensional representations of ultrahigh-dimensional data, e.g., data with thousands/millions of features, has been a major way to enable learning methods to address the curse of dimensionality. However, existing unsupervised representation learning methods mainly focus on preserving the data regularity information and learning the representations independently of subsequent outlier detection methods, which can result in suboptimal and unstable performance of detecting irregularities (i.e., outliers). This paper introduces a ranking model-based framework, called RAMODO, to address this issue. RAMODO unifies representation learning and outlier detection to learn low-dimensional representations that are tailored for a state-of-the-art outlier detection approach - the random distance-based approach. This customized learning yields more optimal and stable representations for the targeted outlier detectors. Additionally, RAMODO can leverage little labeled data as prior knowledge to learn more expressive and application-relevant representations. We instantiate RAMODO to an efficient method called REPEN to demonstrate the performance of RAMODO. Extensive empirical results on eight real-world ultrahigh dimensional data sets show that REPEN (i) enables a random distance-based detector to obtain significantly better AUC performance and two orders of magnitude speedup; (ii) performs substantially better and more stably than four state-of-the-art representation learning methods; and (iii) leverages less than 1% labeled data to achieve up to 32% AUC improvement.

KW - Dimension Reduction

KW - Outlier Detection

KW - Representation Learning

KW - Ultrahigh-dimensional Data

UR - http://www.scopus.com/inward/record.url?scp=85051558867&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85051558867&partnerID=8YFLogxK

U2 - 10.1145/3219819.3220042

DO - 10.1145/3219819.3220042

M3 - Conference contribution

SN - 9781450355520

SP - 2041

EP - 2050

BT - KDD 2018 - Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining

PB - Association for Computing Machinery

ER -