Batch Singular Value Polarization and Weighted Semantic Augmentation for Universal Domain Adaptation

Wang Ziqi, Wei Wang, Chao Huang, Jie Wen, Cong Wang
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:52361-52371, 2024.

Abstract

As a more challenging domain adaptation setting, universal domain adaptation (UniDA) introduces category shift on top of domain shift, which needs to identify unknown category in the target domain and avoid misclassifying target samples into source private categories. To this end, we propose a novel UniDA approach named Batch Singular value Polarization and Weighted Semantic Augmentation (BSP-WSA). Specifically, we adopt an adversarial classifier to identify the target unknown category and align feature distributions between the two domains. Then, we propose to perform SVD on the classifier’s outputs to maximize larger singular values while minimizing those smaller ones, which could prevent target samples from being wrongly assigned to source private classes. To better bridge the domain gap, we propose a weighted semantic augmentation approach for UniDA to generate data on common categories between the two domains. Extensive experiments on three benchmarks demonstrate that BSP-WSA could outperform existing state-of-the-art UniDA approaches.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-ziqi24a, title = {Batch Singular Value Polarization and Weighted Semantic Augmentation for Universal Domain Adaptation}, author = {Ziqi, Wang and Wang, Wei and Huang, Chao and Wen, Jie and Wang, Cong}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {52361--52371}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/ziqi24a/ziqi24a.pdf}, url = {https://proceedings.mlr.press/v235/ziqi24a.html}, abstract = {As a more challenging domain adaptation setting, universal domain adaptation (UniDA) introduces category shift on top of domain shift, which needs to identify unknown category in the target domain and avoid misclassifying target samples into source private categories. To this end, we propose a novel UniDA approach named Batch Singular value Polarization and Weighted Semantic Augmentation (BSP-WSA). Specifically, we adopt an adversarial classifier to identify the target unknown category and align feature distributions between the two domains. Then, we propose to perform SVD on the classifier’s outputs to maximize larger singular values while minimizing those smaller ones, which could prevent target samples from being wrongly assigned to source private classes. To better bridge the domain gap, we propose a weighted semantic augmentation approach for UniDA to generate data on common categories between the two domains. Extensive experiments on three benchmarks demonstrate that BSP-WSA could outperform existing state-of-the-art UniDA approaches.} }
Endnote
%0 Conference Paper %T Batch Singular Value Polarization and Weighted Semantic Augmentation for Universal Domain Adaptation %A Wang Ziqi %A Wei Wang %A Chao Huang %A Jie Wen %A Cong Wang %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-ziqi24a %I PMLR %P 52361--52371 %U https://proceedings.mlr.press/v235/ziqi24a.html %V 235 %X As a more challenging domain adaptation setting, universal domain adaptation (UniDA) introduces category shift on top of domain shift, which needs to identify unknown category in the target domain and avoid misclassifying target samples into source private categories. To this end, we propose a novel UniDA approach named Batch Singular value Polarization and Weighted Semantic Augmentation (BSP-WSA). Specifically, we adopt an adversarial classifier to identify the target unknown category and align feature distributions between the two domains. Then, we propose to perform SVD on the classifier’s outputs to maximize larger singular values while minimizing those smaller ones, which could prevent target samples from being wrongly assigned to source private classes. To better bridge the domain gap, we propose a weighted semantic augmentation approach for UniDA to generate data on common categories between the two domains. Extensive experiments on three benchmarks demonstrate that BSP-WSA could outperform existing state-of-the-art UniDA approaches.
APA
Ziqi, W., Wang, W., Huang, C., Wen, J. & Wang, C.. (2024). Batch Singular Value Polarization and Weighted Semantic Augmentation for Universal Domain Adaptation. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:52361-52371 Available from https://proceedings.mlr.press/v235/ziqi24a.html.

Related Material