Recovering manifold representations via unsupervised meta-learning
Manifold representation learning holds great promise for theoretical understanding and characterization of deep neural networks' behaviors through the lens of geometries. However, data scarcity remains a major challenge in manifold analysis especially for data and applications with real-world c...
Saved in:
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2025-01-01
|
| Series: | Frontiers in Computer Science |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fcomp.2024.1255517/full |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1841527909329666048 |
|---|---|
| author | Yunye Gong Jiachen Yao Ruyi Lian Xiao Lin Chao Chen Ajay Divakaran Yi Yao |
| author_facet | Yunye Gong Jiachen Yao Ruyi Lian Xiao Lin Chao Chen Ajay Divakaran Yi Yao |
| author_sort | Yunye Gong |
| collection | DOAJ |
| description | Manifold representation learning holds great promise for theoretical understanding and characterization of deep neural networks' behaviors through the lens of geometries. However, data scarcity remains a major challenge in manifold analysis especially for data and applications with real-world complexity. To address this issue, we propose manifold representation meta-learning (MRML) based on autoencoders to recover the underlying manifold structures without uniformly or densely sampled data. Specifically, we adopt episodic training, following model agnostic meta-learning, to meta-learn autoencoders that are generalizable to unseen samples specifically corresponding to regions with low-sampling density. We demonstrate the effectiveness of MRML via empirical experiments on LineMOD, a dataset curated for 6-D object pose estimation. We also apply topological metrics based on persistent homology and neighborhood graphs for quantitative assessment of manifolds reconstructed by MRML. In comparison to state-of-the-art baselines, our proposed approach demonstrates improved manifold reconstruction better matching the data manifold by preserving prominent topological features and relative proximity of samples. |
| format | Article |
| id | doaj-art-8740c750336649d891814b55e0b2a30f |
| institution | Kabale University |
| issn | 2624-9898 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Computer Science |
| spelling | doaj-art-8740c750336649d891814b55e0b2a30f2025-01-15T06:10:24ZengFrontiers Media S.A.Frontiers in Computer Science2624-98982025-01-01610.3389/fcomp.2024.12555171255517Recovering manifold representations via unsupervised meta-learningYunye Gong0Jiachen Yao1Ruyi Lian2Xiao Lin3Chao Chen4Ajay Divakaran5Yi Yao6Center for Vision Technologies, SRI International, Princeton, NJ, United StatesDepartment of Computer Science, Stony Brook University, Stony Brook, NY, United StatesDepartment of Computer Science, Stony Brook University, Stony Brook, NY, United StatesCenter for Vision Technologies, SRI International, Princeton, NJ, United StatesDepartment of Computer Science, Stony Brook University, Stony Brook, NY, United StatesCenter for Vision Technologies, SRI International, Princeton, NJ, United StatesCenter for Vision Technologies, SRI International, Princeton, NJ, United StatesManifold representation learning holds great promise for theoretical understanding and characterization of deep neural networks' behaviors through the lens of geometries. However, data scarcity remains a major challenge in manifold analysis especially for data and applications with real-world complexity. To address this issue, we propose manifold representation meta-learning (MRML) based on autoencoders to recover the underlying manifold structures without uniformly or densely sampled data. Specifically, we adopt episodic training, following model agnostic meta-learning, to meta-learn autoencoders that are generalizable to unseen samples specifically corresponding to regions with low-sampling density. We demonstrate the effectiveness of MRML via empirical experiments on LineMOD, a dataset curated for 6-D object pose estimation. We also apply topological metrics based on persistent homology and neighborhood graphs for quantitative assessment of manifolds reconstructed by MRML. In comparison to state-of-the-art baselines, our proposed approach demonstrates improved manifold reconstruction better matching the data manifold by preserving prominent topological features and relative proximity of samples.https://www.frontiersin.org/articles/10.3389/fcomp.2024.1255517/fullmanifold representation learningautoencodermeta-learningpersistent homologydata scarcity |
| spellingShingle | Yunye Gong Jiachen Yao Ruyi Lian Xiao Lin Chao Chen Ajay Divakaran Yi Yao Recovering manifold representations via unsupervised meta-learning Frontiers in Computer Science manifold representation learning autoencoder meta-learning persistent homology data scarcity |
| title | Recovering manifold representations via unsupervised meta-learning |
| title_full | Recovering manifold representations via unsupervised meta-learning |
| title_fullStr | Recovering manifold representations via unsupervised meta-learning |
| title_full_unstemmed | Recovering manifold representations via unsupervised meta-learning |
| title_short | Recovering manifold representations via unsupervised meta-learning |
| title_sort | recovering manifold representations via unsupervised meta learning |
| topic | manifold representation learning autoencoder meta-learning persistent homology data scarcity |
| url | https://www.frontiersin.org/articles/10.3389/fcomp.2024.1255517/full |
| work_keys_str_mv | AT yunyegong recoveringmanifoldrepresentationsviaunsupervisedmetalearning AT jiachenyao recoveringmanifoldrepresentationsviaunsupervisedmetalearning AT ruyilian recoveringmanifoldrepresentationsviaunsupervisedmetalearning AT xiaolin recoveringmanifoldrepresentationsviaunsupervisedmetalearning AT chaochen recoveringmanifoldrepresentationsviaunsupervisedmetalearning AT ajaydivakaran recoveringmanifoldrepresentationsviaunsupervisedmetalearning AT yiyao recoveringmanifoldrepresentationsviaunsupervisedmetalearning |