Class Incremental Learning With Large Domain Shift
We address an important and practical problem facing deep-learning-based image classification: class incremental learning with a large domain shift. Most previous efforts on class incremental learning focus on one aspect of the problem, i.e., learning to classify additional new classes (with a littl...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10759656/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846128481873690624 |
|---|---|
| author | Kamin Lee Hyoeun Kim Geunjae Choi Hyejeong Jeon Nojun Kwak |
| author_facet | Kamin Lee Hyoeun Kim Geunjae Choi Hyejeong Jeon Nojun Kwak |
| author_sort | Kamin Lee |
| collection | DOAJ |
| description | We address an important and practical problem facing deep-learning-based image classification: class incremental learning with a large domain shift. Most previous efforts on class incremental learning focus on one aspect of the problem, i.e., learning to classify additional new classes (with a little shift). However, in the real world, when new classes are added, the domain changes simultaneously (with a large domain shift). To obtain a model that is robust to these situations, we need to consider incrementally learning not only new labels but also domain-shifted labels. We target a continual and simultaneous shift of class and domain distribution and propose a new incremental learning method named Momentum Contrastive learning enhancing Orthogonality of Negative pairs (MoCo-ON). We employ a momentum encoder framework augmented with rehearsal memory to mitigate the risk of forgetting while leveraging contrastive learning to extract versatile features capable of adapting to the progressively shifting domain. Specifically, when training with a knowledge distillation loss, we introduce a novel supervised contrastive loss designed to closely embed positive pairs of the same class, even in the presence of a substantial domain gap. Additionally, we leverage feature embedding from momentum encoders for exemplar selection, aiming to mitigate the risk of forgetting previously acquired information from earlier tasks. We conduct comprehensive experiments involving inter-domain shifted class incremental learning scenarios using widely adopted datasets commonly employed for studying domain generalization in image classification. Our proposed method consistently outperforms other methods by a significant margin. |
| format | Article |
| id | doaj-art-9223d2d8b8c0406a8e57f711a00f7779 |
| institution | Kabale University |
| issn | 2169-3536 |
| language | English |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-9223d2d8b8c0406a8e57f711a00f77792024-12-11T00:06:07ZengIEEEIEEE Access2169-35362024-01-011218356418358010.1109/ACCESS.2024.350428710759656Class Incremental Learning With Large Domain ShiftKamin Lee0https://orcid.org/0009-0005-4608-147XHyoeun Kim1https://orcid.org/0009-0009-6442-9918Geunjae Choi2https://orcid.org/0009-0003-6502-8207Hyejeong Jeon3https://orcid.org/0009-0004-4975-5065Nojun Kwak4https://orcid.org/0000-0002-1792-0327Graduate School of Convergence Science and Technology, Seoul National University, Gwanak, Seoul, Republic of KoreaLG Electronics, Seocho, Seoul, Republic of KoreaGraduate School of Convergence Science and Technology, Seoul National University, Gwanak, Seoul, Republic of KoreaLG Electronics, Seocho, Seoul, Republic of KoreaGraduate School of Convergence Science and Technology, Seoul National University, Gwanak, Seoul, Republic of KoreaWe address an important and practical problem facing deep-learning-based image classification: class incremental learning with a large domain shift. Most previous efforts on class incremental learning focus on one aspect of the problem, i.e., learning to classify additional new classes (with a little shift). However, in the real world, when new classes are added, the domain changes simultaneously (with a large domain shift). To obtain a model that is robust to these situations, we need to consider incrementally learning not only new labels but also domain-shifted labels. We target a continual and simultaneous shift of class and domain distribution and propose a new incremental learning method named Momentum Contrastive learning enhancing Orthogonality of Negative pairs (MoCo-ON). We employ a momentum encoder framework augmented with rehearsal memory to mitigate the risk of forgetting while leveraging contrastive learning to extract versatile features capable of adapting to the progressively shifting domain. Specifically, when training with a knowledge distillation loss, we introduce a novel supervised contrastive loss designed to closely embed positive pairs of the same class, even in the presence of a substantial domain gap. Additionally, we leverage feature embedding from momentum encoders for exemplar selection, aiming to mitigate the risk of forgetting previously acquired information from earlier tasks. We conduct comprehensive experiments involving inter-domain shifted class incremental learning scenarios using widely adopted datasets commonly employed for studying domain generalization in image classification. Our proposed method consistently outperforms other methods by a significant margin.https://ieeexplore.ieee.org/document/10759656/Class incremental learningcontinual learningdomain incremental learningincremental learninglife-long learning |
| spellingShingle | Kamin Lee Hyoeun Kim Geunjae Choi Hyejeong Jeon Nojun Kwak Class Incremental Learning With Large Domain Shift IEEE Access Class incremental learning continual learning domain incremental learning incremental learning life-long learning |
| title | Class Incremental Learning With Large Domain Shift |
| title_full | Class Incremental Learning With Large Domain Shift |
| title_fullStr | Class Incremental Learning With Large Domain Shift |
| title_full_unstemmed | Class Incremental Learning With Large Domain Shift |
| title_short | Class Incremental Learning With Large Domain Shift |
| title_sort | class incremental learning with large domain shift |
| topic | Class incremental learning continual learning domain incremental learning incremental learning life-long learning |
| url | https://ieeexplore.ieee.org/document/10759656/ |
| work_keys_str_mv | AT kaminlee classincrementallearningwithlargedomainshift AT hyoeunkim classincrementallearningwithlargedomainshift AT geunjaechoi classincrementallearningwithlargedomainshift AT hyejeongjeon classincrementallearningwithlargedomainshift AT nojunkwak classincrementallearningwithlargedomainshift |