Exploring Adversarial Robustness of Deep Metric Learning
Deep Metric Learning (DML), a widely-used technique, involves learning a distance metric between pairs of samples. DML uses deep neural architectures to learn semantic embeddings of the input, where the distance between similar examples is small while dissimilar ones are far apart. Although the unde...
Main Authors: | , , , , |
---|---|
Format: | Article in Journal/Newspaper |
Language: | unknown |
Published: |
arXiv
2021
|
Subjects: | |
Online Access: | https://dx.doi.org/10.48550/arxiv.2102.07265 https://arxiv.org/abs/2102.07265 |
id |
ftdatacite:10.48550/arxiv.2102.07265 |
---|---|
record_format |
openpolar |
spelling |
ftdatacite:10.48550/arxiv.2102.07265 2023-05-15T16:01:09+02:00 Exploring Adversarial Robustness of Deep Metric Learning Panum, Thomas Kobber Wang, Zi Kan, Pengyu Fernandes, Earlence Jha, Somesh 2021 https://dx.doi.org/10.48550/arxiv.2102.07265 https://arxiv.org/abs/2102.07265 unknown arXiv arXiv.org perpetual, non-exclusive license http://arxiv.org/licenses/nonexclusive-distrib/1.0/ Machine Learning cs.LG Artificial Intelligence cs.AI FOS Computer and information sciences Article CreativeWork article Preprint 2021 ftdatacite https://doi.org/10.48550/arxiv.2102.07265 2022-03-10T14:51:56Z Deep Metric Learning (DML), a widely-used technique, involves learning a distance metric between pairs of samples. DML uses deep neural architectures to learn semantic embeddings of the input, where the distance between similar examples is small while dissimilar ones are far apart. Although the underlying neural networks produce good accuracy on naturally occurring samples, they are vulnerable to adversarially-perturbed samples that reduce performance. We take a first step towards training robust DML models and tackle the primary challenge of the metric losses being dependent on the samples in a mini-batch, unlike standard losses that only depend on the specific input-output pair. We analyze this dependence effect and contribute a robust optimization formulation. Using experiments on three commonly-used DML datasets, we demonstrate 5-76 fold increases in adversarial accuracy, and outperform an existing DML model that sought out to be robust. 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 Artificial Intelligence cs.AI FOS Computer and information sciences |
spellingShingle |
Machine Learning cs.LG Artificial Intelligence cs.AI FOS Computer and information sciences Panum, Thomas Kobber Wang, Zi Kan, Pengyu Fernandes, Earlence Jha, Somesh Exploring Adversarial Robustness of Deep Metric Learning |
topic_facet |
Machine Learning cs.LG Artificial Intelligence cs.AI FOS Computer and information sciences |
description |
Deep Metric Learning (DML), a widely-used technique, involves learning a distance metric between pairs of samples. DML uses deep neural architectures to learn semantic embeddings of the input, where the distance between similar examples is small while dissimilar ones are far apart. Although the underlying neural networks produce good accuracy on naturally occurring samples, they are vulnerable to adversarially-perturbed samples that reduce performance. We take a first step towards training robust DML models and tackle the primary challenge of the metric losses being dependent on the samples in a mini-batch, unlike standard losses that only depend on the specific input-output pair. We analyze this dependence effect and contribute a robust optimization formulation. Using experiments on three commonly-used DML datasets, we demonstrate 5-76 fold increases in adversarial accuracy, and outperform an existing DML model that sought out to be robust. |
format |
Article in Journal/Newspaper |
author |
Panum, Thomas Kobber Wang, Zi Kan, Pengyu Fernandes, Earlence Jha, Somesh |
author_facet |
Panum, Thomas Kobber Wang, Zi Kan, Pengyu Fernandes, Earlence Jha, Somesh |
author_sort |
Panum, Thomas Kobber |
title |
Exploring Adversarial Robustness of Deep Metric Learning |
title_short |
Exploring Adversarial Robustness of Deep Metric Learning |
title_full |
Exploring Adversarial Robustness of Deep Metric Learning |
title_fullStr |
Exploring Adversarial Robustness of Deep Metric Learning |
title_full_unstemmed |
Exploring Adversarial Robustness of Deep Metric Learning |
title_sort |
exploring adversarial robustness of deep metric learning |
publisher |
arXiv |
publishDate |
2021 |
url |
https://dx.doi.org/10.48550/arxiv.2102.07265 https://arxiv.org/abs/2102.07265 |
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.2102.07265 |
_version_ |
1766397135332311040 |