Beyond UDA: Examining Temporal and Frequency Representations in Time Series Transfer

Ching Chieh Tsao, Fang-Yi Su, Jung-Hsien Chiang
Proceedings of the 17th Asian Conference on Machine Learning, PMLR 304:303-318, 2025.

Abstract

In time-series unsupervised domain adaptation (UDA), the adaptation between temporal and frequency domain features has been relatively underexplored. To address this gap, we conduct a comprehensive series of experiments to revisit the roles of these domains in UDA. Our findings reveal that the temporal domain contains more diverse features, offering higher discriminability, while the frequency domain is more domain-invariant, providing better transferability. Combining the strengths of both domains, we propose TF-DAN, a UDA framework that synergistically integrates temporal and frequency domain features. TF-DAN enhances feature extraction and captures subtle, class-specific features without relying on traditional alignment strategies. By utilizing simple hyperparameter adjustments and using frequency embeddings from the source domain as reference points for domain adaptation, TF-DAN achieves nearly a 10% improvement across five benchmark datasets in time-series UDA. This research highlights the unique strengths of both domains and marks a paradigm shift in UDA methods, showcasing TF-DAN’s robust performance in real-world applications. Codes can be found in the additional material folder.

Cite this Paper

BibTeX
@InProceedings{pmlr-v304-tsao25a, title = {Beyond UDA: Examining Temporal and Frequency Representations in Time Series Transfer}, author = {Tsao, Ching Chieh and Su, Fang-Yi and Chiang, Jung-Hsien}, booktitle = {Proceedings of the 17th Asian Conference on Machine Learning}, pages = {303--318}, year = {2025}, editor = {Lee, Hung-yi and Liu, Tongliang}, volume = {304}, series = {Proceedings of Machine Learning Research}, month = {09--12 Dec}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v304/main/assets/tsao25a/tsao25a.pdf}, url = {https://proceedings.mlr.press/v304/tsao25a.html}, abstract = {In time-series unsupervised domain adaptation (UDA), the adaptation between temporal and frequency domain features has been relatively underexplored. To address this gap, we conduct a comprehensive series of experiments to revisit the roles of these domains in UDA. Our findings reveal that the temporal domain contains more diverse features, offering higher discriminability, while the frequency domain is more domain-invariant, providing better transferability. Combining the strengths of both domains, we propose TF-DAN, a UDA framework that synergistically integrates temporal and frequency domain features. TF-DAN enhances feature extraction and captures subtle, class-specific features without relying on traditional alignment strategies. By utilizing simple hyperparameter adjustments and using frequency embeddings from the source domain as reference points for domain adaptation, TF-DAN achieves nearly a 10% improvement across five benchmark datasets in time-series UDA. This research highlights the unique strengths of both domains and marks a paradigm shift in UDA methods, showcasing TF-DAN’s robust performance in real-world applications. Codes can be found in the additional material folder.} }
Endnote
%0 Conference Paper %T Beyond UDA: Examining Temporal and Frequency Representations in Time Series Transfer %A Ching Chieh Tsao %A Fang-Yi Su %A Jung-Hsien Chiang %B Proceedings of the 17th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Hung-yi Lee %E Tongliang Liu %F pmlr-v304-tsao25a %I PMLR %P 303--318 %U https://proceedings.mlr.press/v304/tsao25a.html %V 304 %X In time-series unsupervised domain adaptation (UDA), the adaptation between temporal and frequency domain features has been relatively underexplored. To address this gap, we conduct a comprehensive series of experiments to revisit the roles of these domains in UDA. Our findings reveal that the temporal domain contains more diverse features, offering higher discriminability, while the frequency domain is more domain-invariant, providing better transferability. Combining the strengths of both domains, we propose TF-DAN, a UDA framework that synergistically integrates temporal and frequency domain features. TF-DAN enhances feature extraction and captures subtle, class-specific features without relying on traditional alignment strategies. By utilizing simple hyperparameter adjustments and using frequency embeddings from the source domain as reference points for domain adaptation, TF-DAN achieves nearly a 10% improvement across five benchmark datasets in time-series UDA. This research highlights the unique strengths of both domains and marks a paradigm shift in UDA methods, showcasing TF-DAN’s robust performance in real-world applications. Codes can be found in the additional material folder.
APA
Tsao, C.C., Su, F. & Chiang, J.. (2025). Beyond UDA: Examining Temporal and Frequency Representations in Time Series Transfer. Proceedings of the 17th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 304:303-318 Available from https://proceedings.mlr.press/v304/tsao25a.html.

Related Material