On Balancing Bias and Variance in Unsupervised Multi-Source-Free Domain Adaptation

Maohao Shen, Yuheng Bu, Gregory W. Wornell
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:30976-30991, 2023.

Abstract

Due to privacy, storage, and other constraints, there is a growing need for unsupervised domain adaptation techniques in machine learning that do not require access to the data used to train a collection of source models. Existing methods for multi-source-free domain adaptation (MSFDA) typically train a target model using pseudo-labeled data produced by the source models, which focus on improving the pseudo-labeling techniques or proposing new training objectives. Instead, we aim to analyze the fundamental limits of MSFDA. In particular, we develop an information-theoretic bound on the generalization error of the resulting target model, which illustrates an inherent bias-variance trade-off. We then provide insights on how to balance this trade-off from three perspectives, including domain aggregation, selective pseudo-labeling, and joint feature alignment, which leads to the design of novel algorithms. Experiments on multiple datasets validate our theoretical analysis and demonstrate the state-of-art performance of the proposed algorithm, especially on some of the most challenging datasets, including Office-Home and DomainNet.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-shen23b, title = {On Balancing Bias and Variance in Unsupervised Multi-Source-Free Domain Adaptation}, author = {Shen, Maohao and Bu, Yuheng and Wornell, Gregory W.}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {30976--30991}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/shen23b/shen23b.pdf}, url = {https://proceedings.mlr.press/v202/shen23b.html}, abstract = {Due to privacy, storage, and other constraints, there is a growing need for unsupervised domain adaptation techniques in machine learning that do not require access to the data used to train a collection of source models. Existing methods for multi-source-free domain adaptation (MSFDA) typically train a target model using pseudo-labeled data produced by the source models, which focus on improving the pseudo-labeling techniques or proposing new training objectives. Instead, we aim to analyze the fundamental limits of MSFDA. In particular, we develop an information-theoretic bound on the generalization error of the resulting target model, which illustrates an inherent bias-variance trade-off. We then provide insights on how to balance this trade-off from three perspectives, including domain aggregation, selective pseudo-labeling, and joint feature alignment, which leads to the design of novel algorithms. Experiments on multiple datasets validate our theoretical analysis and demonstrate the state-of-art performance of the proposed algorithm, especially on some of the most challenging datasets, including Office-Home and DomainNet.} }
Endnote
%0 Conference Paper %T On Balancing Bias and Variance in Unsupervised Multi-Source-Free Domain Adaptation %A Maohao Shen %A Yuheng Bu %A Gregory W. Wornell %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-shen23b %I PMLR %P 30976--30991 %U https://proceedings.mlr.press/v202/shen23b.html %V 202 %X Due to privacy, storage, and other constraints, there is a growing need for unsupervised domain adaptation techniques in machine learning that do not require access to the data used to train a collection of source models. Existing methods for multi-source-free domain adaptation (MSFDA) typically train a target model using pseudo-labeled data produced by the source models, which focus on improving the pseudo-labeling techniques or proposing new training objectives. Instead, we aim to analyze the fundamental limits of MSFDA. In particular, we develop an information-theoretic bound on the generalization error of the resulting target model, which illustrates an inherent bias-variance trade-off. We then provide insights on how to balance this trade-off from three perspectives, including domain aggregation, selective pseudo-labeling, and joint feature alignment, which leads to the design of novel algorithms. Experiments on multiple datasets validate our theoretical analysis and demonstrate the state-of-art performance of the proposed algorithm, especially on some of the most challenging datasets, including Office-Home and DomainNet.
APA
Shen, M., Bu, Y. & Wornell, G.W.. (2023). On Balancing Bias and Variance in Unsupervised Multi-Source-Free Domain Adaptation. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:30976-30991 Available from https://proceedings.mlr.press/v202/shen23b.html.

Related Material