TY - GEN
T1 - Contrastively Smoothed Class Alignment for Unsupervised Domain Adaptation
AU - Dai, Shuyang
AU - Cheng, Yu
AU - Zhang, Yizhe
AU - Gan, Zhe
AU - Liu, Jingjing
AU - Carin, Lawrence
N1 - Publisher Copyright:
© 2021, Springer Nature Switzerland AG.
PY - 2021
Y1 - 2021
N2 - Recent unsupervised approaches to domain adaptation primarily focus on minimizing the gap between the source and the target domains through refining the feature generator, in order to learn a better alignment between the two domains. This minimization can be achieved via a domain classifier to detect target-domain features that are divergent from source-domain features. However, when optimizing via such domain-classification discrepancy, ambiguous target samples that are not smoothly distributed on the low-dimensional data manifold are often missed. To solve this issue, we propose a novel Contrastively Smoothed Class Alignment (CoSCA) model, that explicitly incorporates both intra- and inter-class domain discrepancy to better align ambiguous target samples with the source domain. CoSCA estimates the underlying label hypothesis of target samples, and simultaneously adapts their feature representations by optimizing a proposed contrastive loss. In addition, Maximum Mean Discrepancy (MMD) is utilized to directly match features between source and target samples for better global alignment. Experiments on several benchmark datasets demonstrate that CoSCAoutperforms state-of-the-art approaches for unsupervised domain adaptation by producing more discriminative features.
AB - Recent unsupervised approaches to domain adaptation primarily focus on minimizing the gap between the source and the target domains through refining the feature generator, in order to learn a better alignment between the two domains. This minimization can be achieved via a domain classifier to detect target-domain features that are divergent from source-domain features. However, when optimizing via such domain-classification discrepancy, ambiguous target samples that are not smoothly distributed on the low-dimensional data manifold are often missed. To solve this issue, we propose a novel Contrastively Smoothed Class Alignment (CoSCA) model, that explicitly incorporates both intra- and inter-class domain discrepancy to better align ambiguous target samples with the source domain. CoSCA estimates the underlying label hypothesis of target samples, and simultaneously adapts their feature representations by optimizing a proposed contrastive loss. In addition, Maximum Mean Discrepancy (MMD) is utilized to directly match features between source and target samples for better global alignment. Experiments on several benchmark datasets demonstrate that CoSCAoutperforms state-of-the-art approaches for unsupervised domain adaptation by producing more discriminative features.
UR - http://www.scopus.com/inward/record.url?scp=85103246335&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-69538-5_17
DO - 10.1007/978-3-030-69538-5_17
M3 - Conference contribution
AN - SCOPUS:85103246335
SN - 9783030695378
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 268
EP - 283
BT - Computer Vision – ACCV 2020 - 15th Asian Conference on Computer Vision, 2020, Revised Selected Papers
A2 - Ishikawa, Hiroshi
A2 - Liu, Cheng-Lin
A2 - Pajdla, Tomas
A2 - Shi, Jianbo
PB - Springer Science and Business Media Deutschland GmbH
T2 - 15th Asian Conference on Computer Vision, ACCV 2020
Y2 - 30 November 2020 through 4 December 2020
ER -