Unsupervised Anomaly Detection via Deep Metric Learning with End-to-End Optimization
We investigate unsupervised anomaly detection for high-dimensional data and introduce a deep metric learning (DML) based framework. In particular, we learn a distance metric through a deep neural network. Through this metric, we project the data into the metric space that better separates the anomal...
Main Authors: | , |
---|---|
Format: | Article in Journal/Newspaper |
Language: | unknown |
Published: |
arXiv
2020
|
Subjects: | |
Online Access: | https://dx.doi.org/10.48550/arxiv.2005.05865 https://arxiv.org/abs/2005.05865 |
id |
ftdatacite:10.48550/arxiv.2005.05865 |
---|---|
record_format |
openpolar |
spelling |
ftdatacite:10.48550/arxiv.2005.05865 2023-05-15T16:01:38+02:00 Unsupervised Anomaly Detection via Deep Metric Learning with End-to-End Optimization Yilmaz, Selim F. Kozat, Suleyman S. 2020 https://dx.doi.org/10.48550/arxiv.2005.05865 https://arxiv.org/abs/2005.05865 unknown arXiv arXiv.org perpetual, non-exclusive license http://arxiv.org/licenses/nonexclusive-distrib/1.0/ Machine Learning cs.LG Machine Learning stat.ML FOS Computer and information sciences Article CreativeWork article Preprint 2020 ftdatacite https://doi.org/10.48550/arxiv.2005.05865 2022-03-10T15:44:11Z We investigate unsupervised anomaly detection for high-dimensional data and introduce a deep metric learning (DML) based framework. In particular, we learn a distance metric through a deep neural network. Through this metric, we project the data into the metric space that better separates the anomalies from the normal data and reduces the effect of the curse of dimensionality for high-dimensional data. We present a novel data distillation method through self-supervision to remedy the conventional practice of assuming all data as normal. We also employ the hard mining technique from the DML literature. We show these components improve the performance of our model and significantly reduce the running time. Through an extensive set of experiments on the 14 real-world datasets, our method demonstrates significant performance gains compared to the state-of-the-art unsupervised anomaly detection methods, e.g., an absolute improvement between 4.44% and 11.74% on the average over the 14 datasets. Furthermore, we share the source code of our method on Github to facilitate further research. : 11 pages, 3 figures Article in Journal/Newspaper DML DataCite Metadata Store (German National Library of Science and Technology) |
institution |
Open Polar |
collection |
DataCite Metadata Store (German National Library of Science and Technology) |
op_collection_id |
ftdatacite |
language |
unknown |
topic |
Machine Learning cs.LG Machine Learning stat.ML FOS Computer and information sciences |
spellingShingle |
Machine Learning cs.LG Machine Learning stat.ML FOS Computer and information sciences Yilmaz, Selim F. Kozat, Suleyman S. Unsupervised Anomaly Detection via Deep Metric Learning with End-to-End Optimization |
topic_facet |
Machine Learning cs.LG Machine Learning stat.ML FOS Computer and information sciences |
description |
We investigate unsupervised anomaly detection for high-dimensional data and introduce a deep metric learning (DML) based framework. In particular, we learn a distance metric through a deep neural network. Through this metric, we project the data into the metric space that better separates the anomalies from the normal data and reduces the effect of the curse of dimensionality for high-dimensional data. We present a novel data distillation method through self-supervision to remedy the conventional practice of assuming all data as normal. We also employ the hard mining technique from the DML literature. We show these components improve the performance of our model and significantly reduce the running time. Through an extensive set of experiments on the 14 real-world datasets, our method demonstrates significant performance gains compared to the state-of-the-art unsupervised anomaly detection methods, e.g., an absolute improvement between 4.44% and 11.74% on the average over the 14 datasets. Furthermore, we share the source code of our method on Github to facilitate further research. : 11 pages, 3 figures |
format |
Article in Journal/Newspaper |
author |
Yilmaz, Selim F. Kozat, Suleyman S. |
author_facet |
Yilmaz, Selim F. Kozat, Suleyman S. |
author_sort |
Yilmaz, Selim F. |
title |
Unsupervised Anomaly Detection via Deep Metric Learning with End-to-End Optimization |
title_short |
Unsupervised Anomaly Detection via Deep Metric Learning with End-to-End Optimization |
title_full |
Unsupervised Anomaly Detection via Deep Metric Learning with End-to-End Optimization |
title_fullStr |
Unsupervised Anomaly Detection via Deep Metric Learning with End-to-End Optimization |
title_full_unstemmed |
Unsupervised Anomaly Detection via Deep Metric Learning with End-to-End Optimization |
title_sort |
unsupervised anomaly detection via deep metric learning with end-to-end optimization |
publisher |
arXiv |
publishDate |
2020 |
url |
https://dx.doi.org/10.48550/arxiv.2005.05865 https://arxiv.org/abs/2005.05865 |
genre |
DML |
genre_facet |
DML |
op_rights |
arXiv.org perpetual, non-exclusive license http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
op_doi |
https://doi.org/10.48550/arxiv.2005.05865 |
_version_ |
1766397402759036928 |