TY - JOUR
T1 - Domain Alignment Meets Fully Test-Time Adaptation
AU - Thopalli, Kowshik
AU - Turaga, Pavan
AU - Thiagarajan, Jayaraman J.
N1 - Funding Information:
KT’s and PT’s work are supported by the Defense Advanced Research Projects Agency (DARPA) under Agreement No. HR00112290073. JT’s work was supported by the LDRD Program under project 21-ERD-012.
Funding Information:
. This work was performed under the auspices of the U.S. Department of Energy by the Lawrence Liv-ermore National Laboratory under Contract No. DE-AC52-07NA27344, Lawrence Livermore National Security, LLC.
Publisher Copyright:
© 2022 K. Thopalli, P. Turaga & J.J. Thiagarajan.
PY - 2022
Y1 - 2022
N2 - A foundational requirement of a deployed ML model is to generalize to data drawn from a testing distribution that is different from training. A popular solution to this problem is to adapt a pre-trained model to novel domains using only unlabeled data. In this paper, we focus on a challenging variant of this problem, where access to the original source data is restricted. While fully test-time adaptation (FTTA) and unsupervised domain adaptation (UDA) are closely related, the advances in UDA are not readily applicable to TTA, since most UDA methods require access to the source data. Hence, we propose a new approach, CATTAn, that bridges UDA and FTTA, by relaxing the need to access entire source data, through a novel deep subspace alignment strategy. With a minimal overhead of storing the subspace basis set for the source data, CATTAn enables unsupervised alignment between source and target data during adaptation. Through extensive experimental evaluation on multiple 2D and 3D vision benchmarks (ImageNet-C, Office-31, OfficeHome, DomainNet, PointDA-10) and model architectures, we demonstrate significant gains in FTTA performance. Furthermore, we make a number of crucial findings on the utility of the alignment objective even with inherently robust models, pre-trained ViT representations and under low sample availability in the target domain.
AB - A foundational requirement of a deployed ML model is to generalize to data drawn from a testing distribution that is different from training. A popular solution to this problem is to adapt a pre-trained model to novel domains using only unlabeled data. In this paper, we focus on a challenging variant of this problem, where access to the original source data is restricted. While fully test-time adaptation (FTTA) and unsupervised domain adaptation (UDA) are closely related, the advances in UDA are not readily applicable to TTA, since most UDA methods require access to the source data. Hence, we propose a new approach, CATTAn, that bridges UDA and FTTA, by relaxing the need to access entire source data, through a novel deep subspace alignment strategy. With a minimal overhead of storing the subspace basis set for the source data, CATTAn enables unsupervised alignment between source and target data during adaptation. Through extensive experimental evaluation on multiple 2D and 3D vision benchmarks (ImageNet-C, Office-31, OfficeHome, DomainNet, PointDA-10) and model architectures, we demonstrate significant gains in FTTA performance. Furthermore, we make a number of crucial findings on the utility of the alignment objective even with inherently robust models, pre-trained ViT representations and under low sample availability in the target domain.
KW - Domain Shifts
KW - Geometric Alignment
KW - Robustness
KW - Test-Time Adaptation
UR - http://www.scopus.com/inward/record.url?scp=85162191939&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85162191939&partnerID=8YFLogxK
M3 - Conference article
AN - SCOPUS:85162191939
SN - 2640-3498
VL - 189
SP - 1006
EP - 1021
JO - Proceedings of Machine Learning Research
JF - Proceedings of Machine Learning Research
T2 - 14th Asian Conference on Machine Learning, ACML 2022
Y2 - 12 December 2022 through 14 December 2022
ER -