Efficient Test-Time Model Adaptation without Forgetting

Shuaicheng Niu, Jiaxiang Wu, Yifan Zhang, Yaofo Chen, Shijian Zheng, Peilin Zhao, Mingkui Tan
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:16888-16905, 2022.

Abstract

Test-time adaptation provides an effective means of tackling the potential distribution shift between model training and inference, by dynamically updating the model at test time. This area has seen fast progress recently, at the effectiveness of handling test shifts. Nonetheless, prior methods still suffer two key limitations: 1) these methods rely on performing backward computation for each test sample, which takes a considerable amount of time; and 2) these methods focus on improving the performance on out-of-distribution test samples and ignore that the adaptation on test data may result in a catastrophic forgetting issue, \ie, the performance on in-distribution test samples may degrade. To address these issues, we propose an efficient anti-forgetting test-time adaptation (EATA) method. Specifically, we devise a sample-efficient entropy minimization loss to exclude uninformative samples out of backward computation, which improves the overall efficiency and meanwhile boosts the out-of-distribution accuracy. Afterward, we introduce a regularization loss to ensure that critical model weights tend to be preserved during adaptation, thereby alleviating the forgetting issue. Extensive experiments on CIFAR-10-C, ImageNet-C, and ImageNet-R verify the effectiveness and superiority of our EATA.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-niu22a, title = {Efficient Test-Time Model Adaptation without Forgetting}, author = {Niu, Shuaicheng and Wu, Jiaxiang and Zhang, Yifan and Chen, Yaofo and Zheng, Shijian and Zhao, Peilin and Tan, Mingkui}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {16888--16905}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/niu22a/niu22a.pdf}, url = {https://proceedings.mlr.press/v162/niu22a.html}, abstract = {Test-time adaptation provides an effective means of tackling the potential distribution shift between model training and inference, by dynamically updating the model at test time. This area has seen fast progress recently, at the effectiveness of handling test shifts. Nonetheless, prior methods still suffer two key limitations: 1) these methods rely on performing backward computation for each test sample, which takes a considerable amount of time; and 2) these methods focus on improving the performance on out-of-distribution test samples and ignore that the adaptation on test data may result in a catastrophic forgetting issue, \ie, the performance on in-distribution test samples may degrade. To address these issues, we propose an efficient anti-forgetting test-time adaptation (EATA) method. Specifically, we devise a sample-efficient entropy minimization loss to exclude uninformative samples out of backward computation, which improves the overall efficiency and meanwhile boosts the out-of-distribution accuracy. Afterward, we introduce a regularization loss to ensure that critical model weights tend to be preserved during adaptation, thereby alleviating the forgetting issue. Extensive experiments on CIFAR-10-C, ImageNet-C, and ImageNet-R verify the effectiveness and superiority of our EATA.} }
Endnote
%0 Conference Paper %T Efficient Test-Time Model Adaptation without Forgetting %A Shuaicheng Niu %A Jiaxiang Wu %A Yifan Zhang %A Yaofo Chen %A Shijian Zheng %A Peilin Zhao %A Mingkui Tan %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-niu22a %I PMLR %P 16888--16905 %U https://proceedings.mlr.press/v162/niu22a.html %V 162 %X Test-time adaptation provides an effective means of tackling the potential distribution shift between model training and inference, by dynamically updating the model at test time. This area has seen fast progress recently, at the effectiveness of handling test shifts. Nonetheless, prior methods still suffer two key limitations: 1) these methods rely on performing backward computation for each test sample, which takes a considerable amount of time; and 2) these methods focus on improving the performance on out-of-distribution test samples and ignore that the adaptation on test data may result in a catastrophic forgetting issue, \ie, the performance on in-distribution test samples may degrade. To address these issues, we propose an efficient anti-forgetting test-time adaptation (EATA) method. Specifically, we devise a sample-efficient entropy minimization loss to exclude uninformative samples out of backward computation, which improves the overall efficiency and meanwhile boosts the out-of-distribution accuracy. Afterward, we introduce a regularization loss to ensure that critical model weights tend to be preserved during adaptation, thereby alleviating the forgetting issue. Extensive experiments on CIFAR-10-C, ImageNet-C, and ImageNet-R verify the effectiveness and superiority of our EATA.
APA
Niu, S., Wu, J., Zhang, Y., Chen, Y., Zheng, S., Zhao, P. & Tan, M.. (2022). Efficient Test-Time Model Adaptation without Forgetting. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:16888-16905 Available from https://proceedings.mlr.press/v162/niu22a.html.

Related Material