Transferring Knowledge Fragments for Learning Distance Metric from A Heterogeneous Domain
The goal of transfer learning is to improve the performance of target learning task by leveraging information (or transferring knowledge) from other related tasks. In this paper, we examine the problem of transfer distance metric learning (DML), which usually aims to mitigate the label information d...
Main Authors: | , , , |
---|---|
Format: | Article in Journal/Newspaper |
Language: | unknown |
Published: |
arXiv
2019
|
Subjects: | |
Online Access: | https://dx.doi.org/10.48550/arxiv.1904.04061 https://arxiv.org/abs/1904.04061 |
id |
ftdatacite:10.48550/arxiv.1904.04061 |
---|---|
record_format |
openpolar |
spelling |
ftdatacite:10.48550/arxiv.1904.04061 2023-05-15T16:01:23+02:00 Transferring Knowledge Fragments for Learning Distance Metric from A Heterogeneous Domain Luo, Yong Wen, Yonggang Liu, Tongliang Tao, Dacheng 2019 https://dx.doi.org/10.48550/arxiv.1904.04061 https://arxiv.org/abs/1904.04061 unknown arXiv https://dx.doi.org/10.1109/tpami.2018.2824309 arXiv.org perpetual, non-exclusive license http://arxiv.org/licenses/nonexclusive-distrib/1.0/ Machine Learning stat.ML Computer Vision and Pattern Recognition cs.CV Machine Learning cs.LG FOS Computer and information sciences article-journal Article ScholarlyArticle Text 2019 ftdatacite https://doi.org/10.48550/arxiv.1904.04061 https://doi.org/10.1109/tpami.2018.2824309 2022-03-10T16:43:00Z The goal of transfer learning is to improve the performance of target learning task by leveraging information (or transferring knowledge) from other related tasks. In this paper, we examine the problem of transfer distance metric learning (DML), which usually aims to mitigate the label information deficiency issue in the target DML. Most of the current Transfer DML (TDML) methods are not applicable to the scenario where data are drawn from heterogeneous domains. Some existing heterogeneous transfer learning (HTL) approaches can learn target distance metric by usually transforming the samples of source and target domain into a common subspace. However, these approaches lack flexibility in real-world applications, and the learned transformations are often restricted to be linear. This motivates us to develop a general flexible heterogeneous TDML (HTDML) framework. In particular, any (linear/nonlinear) DML algorithms can be employed to learn the source metric beforehand. Then the pre-learned source metric is represented as a set of knowledge fragments to help target metric learning. We show how generalization error in the target domain could be reduced using the proposed transfer strategy, and develop novel algorithm to learn either linear or nonlinear target metric. Extensive experiments on various applications demonstrate the effectiveness of the proposed method. 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 stat.ML Computer Vision and Pattern Recognition cs.CV Machine Learning cs.LG FOS Computer and information sciences |
spellingShingle |
Machine Learning stat.ML Computer Vision and Pattern Recognition cs.CV Machine Learning cs.LG FOS Computer and information sciences Luo, Yong Wen, Yonggang Liu, Tongliang Tao, Dacheng Transferring Knowledge Fragments for Learning Distance Metric from A Heterogeneous Domain |
topic_facet |
Machine Learning stat.ML Computer Vision and Pattern Recognition cs.CV Machine Learning cs.LG FOS Computer and information sciences |
description |
The goal of transfer learning is to improve the performance of target learning task by leveraging information (or transferring knowledge) from other related tasks. In this paper, we examine the problem of transfer distance metric learning (DML), which usually aims to mitigate the label information deficiency issue in the target DML. Most of the current Transfer DML (TDML) methods are not applicable to the scenario where data are drawn from heterogeneous domains. Some existing heterogeneous transfer learning (HTL) approaches can learn target distance metric by usually transforming the samples of source and target domain into a common subspace. However, these approaches lack flexibility in real-world applications, and the learned transformations are often restricted to be linear. This motivates us to develop a general flexible heterogeneous TDML (HTDML) framework. In particular, any (linear/nonlinear) DML algorithms can be employed to learn the source metric beforehand. Then the pre-learned source metric is represented as a set of knowledge fragments to help target metric learning. We show how generalization error in the target domain could be reduced using the proposed transfer strategy, and develop novel algorithm to learn either linear or nonlinear target metric. Extensive experiments on various applications demonstrate the effectiveness of the proposed method. |
format |
Article in Journal/Newspaper |
author |
Luo, Yong Wen, Yonggang Liu, Tongliang Tao, Dacheng |
author_facet |
Luo, Yong Wen, Yonggang Liu, Tongliang Tao, Dacheng |
author_sort |
Luo, Yong |
title |
Transferring Knowledge Fragments for Learning Distance Metric from A Heterogeneous Domain |
title_short |
Transferring Knowledge Fragments for Learning Distance Metric from A Heterogeneous Domain |
title_full |
Transferring Knowledge Fragments for Learning Distance Metric from A Heterogeneous Domain |
title_fullStr |
Transferring Knowledge Fragments for Learning Distance Metric from A Heterogeneous Domain |
title_full_unstemmed |
Transferring Knowledge Fragments for Learning Distance Metric from A Heterogeneous Domain |
title_sort |
transferring knowledge fragments for learning distance metric from a heterogeneous domain |
publisher |
arXiv |
publishDate |
2019 |
url |
https://dx.doi.org/10.48550/arxiv.1904.04061 https://arxiv.org/abs/1904.04061 |
genre |
DML |
genre_facet |
DML |
op_relation |
https://dx.doi.org/10.1109/tpami.2018.2824309 |
op_rights |
arXiv.org perpetual, non-exclusive license http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
op_doi |
https://doi.org/10.48550/arxiv.1904.04061 https://doi.org/10.1109/tpami.2018.2824309 |
_version_ |
1766397275574108160 |