In Search for a Generalizable Method for Source Free Domain Adaptation

Malik Boudiaf, Tom Denton, Bart Van Merrienboer, Vincent Dumoulin, Eleni Triantafillou
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:2914-2931, 2023.

Abstract

Source-free domain adaptation (SFDA) is compelling because it allows adapting an off-the-shelf model to a new domain using only unlabelled data. In this work, we apply existing SFDA techniques to a challenging set of naturally-occurring distribution shifts in bioacoustics, which are very different from the ones commonly studied in computer vision. We find existing methods perform differently relative to each other than observed in vision benchmarks, and sometimes perform worse than no adaptation at all. We propose a new simple method which outperforms the existing methods on our new shifts while exhibiting strong performance on a range of vision datasets. Our findings suggest that existing SFDA methods are not as generalizable as previously thought and that considering diverse modalities can be a useful avenue for designing more robust models.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-boudiaf23a, title = {In Search for a Generalizable Method for Source Free Domain Adaptation}, author = {Boudiaf, Malik and Denton, Tom and Van Merrienboer, Bart and Dumoulin, Vincent and Triantafillou, Eleni}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {2914--2931}, 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/boudiaf23a/boudiaf23a.pdf}, url = {https://proceedings.mlr.press/v202/boudiaf23a.html}, abstract = {Source-free domain adaptation (SFDA) is compelling because it allows adapting an off-the-shelf model to a new domain using only unlabelled data. In this work, we apply existing SFDA techniques to a challenging set of naturally-occurring distribution shifts in bioacoustics, which are very different from the ones commonly studied in computer vision. We find existing methods perform differently relative to each other than observed in vision benchmarks, and sometimes perform worse than no adaptation at all. We propose a new simple method which outperforms the existing methods on our new shifts while exhibiting strong performance on a range of vision datasets. Our findings suggest that existing SFDA methods are not as generalizable as previously thought and that considering diverse modalities can be a useful avenue for designing more robust models.} }
Endnote
%0 Conference Paper %T In Search for a Generalizable Method for Source Free Domain Adaptation %A Malik Boudiaf %A Tom Denton %A Bart Van Merrienboer %A Vincent Dumoulin %A Eleni Triantafillou %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-boudiaf23a %I PMLR %P 2914--2931 %U https://proceedings.mlr.press/v202/boudiaf23a.html %V 202 %X Source-free domain adaptation (SFDA) is compelling because it allows adapting an off-the-shelf model to a new domain using only unlabelled data. In this work, we apply existing SFDA techniques to a challenging set of naturally-occurring distribution shifts in bioacoustics, which are very different from the ones commonly studied in computer vision. We find existing methods perform differently relative to each other than observed in vision benchmarks, and sometimes perform worse than no adaptation at all. We propose a new simple method which outperforms the existing methods on our new shifts while exhibiting strong performance on a range of vision datasets. Our findings suggest that existing SFDA methods are not as generalizable as previously thought and that considering diverse modalities can be a useful avenue for designing more robust models.
APA
Boudiaf, M., Denton, T., Van Merrienboer, B., Dumoulin, V. & Triantafillou, E.. (2023). In Search for a Generalizable Method for Source Free Domain Adaptation. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:2914-2931 Available from https://proceedings.mlr.press/v202/boudiaf23a.html.

Related Material