Geometry Flow-Based Deep Riemannian Metric Learning
Deep metric learning (DML) has achieved great results on visual understanding tasks by seamlessly integrating conventional metric learning with deep neural networks. Existing deep metric learning methods focus on designing pair-based distance loss to decrease intra-class distance while increasing in...
Published in: | IEEE/CAA Journal of Automatica Sinica |
---|---|
Main Authors: | , , , , |
Format: | Report |
Language: | unknown |
Published: |
2023
|
Subjects: | |
Online Access: | http://ir.ia.ac.cn/handle/173211/52376 https://doi.org/10.1109/JAS.2023.123399 |
id |
ftchiacadsccasia:oai:ir.ia.ac.cn:173211/52376 |
---|---|
record_format |
openpolar |
spelling |
ftchiacadsccasia:oai:ir.ia.ac.cn:173211/52376 2023-09-05T13:19:06+02:00 Geometry Flow-Based Deep Riemannian Metric Learning Yangyang Li Chaoqun Fei Chuanqing Wang Hongming Shan Ruqian Lu 2023 http://ir.ia.ac.cn/handle/173211/52376 https://doi.org/10.1109/JAS.2023.123399 unknown IEEE/CAA Journal of Automatica Sinica http://ir.ia.ac.cn/handle/173211/52376 doi:10.1109/JAS.2023.123399 cn.org.cspace.api.content.CopyrightPolicy@68325efe Curvature regularization deep metric learning (DML) embedding learning geometry flow riemannian metric 期刊论文 2023 ftchiacadsccasia https://doi.org/10.1109/JAS.2023.123399 2023-08-11T00:18:36Z Deep metric learning (DML) has achieved great results on visual understanding tasks by seamlessly integrating conventional metric learning with deep neural networks. Existing deep metric learning methods focus on designing pair-based distance loss to decrease intra-class distance while increasing inter-class distance. However, these methods fail to preserve the geometric structure of data in the embedding space, which leads to the spatial structure shift across mini-batches and may slow down the convergence of embedding learning. To alleviate these issues, by assuming that the input data is embedded in a lower-dimensional sub-manifold, we propose a novel deep Riemannian metric learning (DRML) framework that exploits the non-Euclidean geometric structural information. Considering that the curvature information of data measures how much the Riemannian (non-Euclidean) metric deviates from the Euclidean metric, we leverage geometry flow, which is called a geometric evolution equation, to characterize the relation between the Riemannian metric and its curvature. Our DRML not only regularizes the local neighborhoods connection of the embeddings at the hidden layer but also adapts the embeddings to preserve the geometric structure of the data. On several benchmark datasets, the proposed DRML outperforms all existing methods and these results demonstrate its effectiveness. Report DML Institute of Automation: CASIA OpenIR (Chinese Academy of Sciences) IEEE/CAA Journal of Automatica Sinica 10 9 1882 1892 |
institution |
Open Polar |
collection |
Institute of Automation: CASIA OpenIR (Chinese Academy of Sciences) |
op_collection_id |
ftchiacadsccasia |
language |
unknown |
topic |
Curvature regularization deep metric learning (DML) embedding learning geometry flow riemannian metric |
spellingShingle |
Curvature regularization deep metric learning (DML) embedding learning geometry flow riemannian metric Yangyang Li Chaoqun Fei Chuanqing Wang Hongming Shan Ruqian Lu Geometry Flow-Based Deep Riemannian Metric Learning |
topic_facet |
Curvature regularization deep metric learning (DML) embedding learning geometry flow riemannian metric |
description |
Deep metric learning (DML) has achieved great results on visual understanding tasks by seamlessly integrating conventional metric learning with deep neural networks. Existing deep metric learning methods focus on designing pair-based distance loss to decrease intra-class distance while increasing inter-class distance. However, these methods fail to preserve the geometric structure of data in the embedding space, which leads to the spatial structure shift across mini-batches and may slow down the convergence of embedding learning. To alleviate these issues, by assuming that the input data is embedded in a lower-dimensional sub-manifold, we propose a novel deep Riemannian metric learning (DRML) framework that exploits the non-Euclidean geometric structural information. Considering that the curvature information of data measures how much the Riemannian (non-Euclidean) metric deviates from the Euclidean metric, we leverage geometry flow, which is called a geometric evolution equation, to characterize the relation between the Riemannian metric and its curvature. Our DRML not only regularizes the local neighborhoods connection of the embeddings at the hidden layer but also adapts the embeddings to preserve the geometric structure of the data. On several benchmark datasets, the proposed DRML outperforms all existing methods and these results demonstrate its effectiveness. |
format |
Report |
author |
Yangyang Li Chaoqun Fei Chuanqing Wang Hongming Shan Ruqian Lu |
author_facet |
Yangyang Li Chaoqun Fei Chuanqing Wang Hongming Shan Ruqian Lu |
author_sort |
Yangyang Li |
title |
Geometry Flow-Based Deep Riemannian Metric Learning |
title_short |
Geometry Flow-Based Deep Riemannian Metric Learning |
title_full |
Geometry Flow-Based Deep Riemannian Metric Learning |
title_fullStr |
Geometry Flow-Based Deep Riemannian Metric Learning |
title_full_unstemmed |
Geometry Flow-Based Deep Riemannian Metric Learning |
title_sort |
geometry flow-based deep riemannian metric learning |
publishDate |
2023 |
url |
http://ir.ia.ac.cn/handle/173211/52376 https://doi.org/10.1109/JAS.2023.123399 |
genre |
DML |
genre_facet |
DML |
op_relation |
IEEE/CAA Journal of Automatica Sinica http://ir.ia.ac.cn/handle/173211/52376 doi:10.1109/JAS.2023.123399 |
op_rights |
cn.org.cspace.api.content.CopyrightPolicy@68325efe |
op_doi |
https://doi.org/10.1109/JAS.2023.123399 |
container_title |
IEEE/CAA Journal of Automatica Sinica |
container_volume |
10 |
container_issue |
9 |
container_start_page |
1882 |
op_container_end_page |
1892 |
_version_ |
1776199910465470464 |