Abstract
By leveraging data from a fully labeled source domain, unsupervised domain adaptation (UDA) improves classification performance on an unlabeled target domain through explicit discrepancy minimization of data distribution or adversarial learning. As an enhancement, category alignment is involved during adaptation to reinforce target feature discrimination by utilizing model prediction. However, there remain unexplored problems about pseudo-label inaccuracy incurred by wrong category predictions on target domain, and distribution deviation caused by overfitting on source domain. In this paper, we propose a model-agnostic two-stage learning framework, which greatly reduces flawed model predictions using soft pseudo-label strategy and avoids overfitting on source domain with a curriculum learning strategy. Theoretically, it successfully decreases the combined risk in the upper bound of expected error on the target domain. In the first stage, we train a model with distribution alignment-based UDA method to obtain soft semantic label on target domain with rather high confidence. To avoid overfitting on source domain, in the second stage, we propose a curriculum learning strategy to adaptively control the weighting between losses from the two domains so that the focus of the training stage is gradually shifted from source distribution to target distribution with prediction confidence boosted on the target domain. Extensive experiments on two well-known benchmark datasets validate the universal effectiveness of our proposed framework on promoting the performance of the top-ranked UDA algorithms and demonstrate its consistent superior performance.
摘要
无监督域适应利用标签完整的源域数据,通过显式的数据分布差异最小化或对抗学习,提高无标签目标域的分类性能。作为一种增强方法,在域适应过程中会涉及类别对齐,即利用模型预测来加强目标特征识别。此方法存在两个问题:在目标域中,错误的类别预测会导致伪标签不准确;在源域中,过拟合会导致分布偏差。因此本文提出了一种与模型无关的两阶段学习框架,利用软伪标签策略大大减少了错误的模型预测,并利用课程学习策略避免了源域的过拟合。理论上,成功降低目标域预期误差上限的综合风险。在第一阶段,我们使用基于分布对齐的无监督域适应方法训练模型,以获得置信度相当高的目标域软语义标签。为了避免源域的过拟合,在第二阶段,我们提出了一种课程学习策略,以自适应性地控制两个域损失之间的权重,从而使训练阶段的重点逐渐从源域分布转移到目标域分布,并提高目标域的预测置信度。在两个常见基准数据集上进行的广泛实验验证了我们提出的框架在提升排名靠前的无监督域适应算法性能方面的普遍有效性,并证明了其一贯的卓越性能。
Similar content being viewed by others
References
SUN B, FENG J, SAENKO K. Return of frustratingly easy domain adaptation [C]//Thirtieth AAAI Conference on Artificial Intelligence. Phoenix, AZ, USA: AAAI, 2016: 2058–2065.
TORRALBA A, EFROS A A. Unbiased look at dataset bias [C]//CVPR 2011. Colorado Springs, CO, USA: IEEE, 2011: 1521–1528.
ZHU Y C, ZHUANG F Z, WANG J D, et al. Deep subdomain adaptation network for image classification [J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(4): 1713–1722.
CUI S H, WANG S H, ZHUO J B, et al. Gradually vanishing bridge for adversarial domain adaptation [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, WA, USA: IEEE, 2020: 12452–12461.
ZHANG W C, OUYANG W L, LI W, et al. Collaborative and adversarial network for unsupervised domain adaptation [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA: IEEE, 2018: 3801–3809.
KANG G L, JIANG L, YANG Y, et al. Contrastive adaptation network for unsupervised domain adaptation [C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, CA, USA: IEEE, 2019: 4888–4897.
LONG M, CAO Z, WANG J, et al. Conditional adversarial domain adaptation [M]//Advances in neural information processing systems 31. Red Hook: Curran Associates Inc., 2018: 1645–1655.
ZHANG Y, LIU T, LONG M, et al. Bridging theory and algorithm for domain adaptation [C]//36th International Conference on Machine Learning. Long Beach, CA, USA: PMLR, 2019: 7404–7413.
XIAO N, ZHANG L. Dynamic weighted learning for unsupervised domain adaptation [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, TN, USA: IEEE, 2021: 15237–15246.
WEI G Q, LAN C L, ZENG W J, et al. MetaAlign: coordinating domain alignment and classification for unsupervised domain adaptation [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, TN, USA: IEEE, 2021: 16638–16648.
SHARMA A, KALLURI T, CHANDRAKER M. Instance level affinity-based transfer for unsupervised domain adaptation [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, TN, USA: IEEE, 2021: 5357–5367.
ZHONG L, FANG Z, LIU F, et al. How does the combined risk affect the performance of unsupervised domain adaptation approaches? [C]//35th AAAI Conference on Artificial Intelligence. Online: AAAI, 2021: 11079–11087.
LI S, XIE M X, GONG K X, et al. Transferable semantic augmentation for domain adaptation [C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, TN, USA: IEEE, 2021: 11511–11520.
BEN-DAVID S, BLITZER J, CRAMMER K, et al. A theory of learning from different domains [J]. Machine Learning, 2010, 79(1/2): 151–175.
XU M H, ZHANG J, NI B B, et al. Adversarial domain adaptation with domain mixup [J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(4): 6502–6509.
ZHANG Y B, DENG B, TANG H, et al. Unsupervised multi-class domain adaptation: Theory, algorithms, and practice [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2022, 44(5): 2775–2792.
GENG B, TAO D C, XU C. DAML: domain adaptation metric learning [J]. IEEE Transactions on Image Processing, 2011, 20(10): 2980–2989.
LONG M, CAO Y, WANG J, et al. Learning transferable features with deep adaptation networks [C]//32 nd International Conference on Machine Learning. Lille, France: PMLA, 2015: 97–105.
TZENG E, HOFFMAN J, ZHANG N, et al. Deep domain confusion: Maximizing for domain invariance [DB/OL]. (2014-12-10). https://arxiv.org/abs/1412.3474.
ZHANG Y B, TANG H, JIA K, et al. Domain-symmetric networks for adversarial domain adaptation [C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, CA, USA: IEEE, 2019: 5026–5035.
PENG X C, BAI Q X, XIA X D, et al. Moment matching for multi-source domain adaptation [C]//2019 IEEE/CVF International Conference on Computer Vision. Seoul, Korea: IEEE, 2019: 1406–1415.
LI X D, HU Y, ZHENG J H, et al. Central moment discrepancy based domain adaptation for intelligent bearing fault diagnosis [J]. Neurocomputing, 2021, 429: 12–24.
PENG X C, SAENKO K. Synthetic to real adaptation with generative correlation alignment networks [C]//2018 IEEE Winter Conference on Applications of Computer Vision. Lake Tahoe, NV, USA: IEEE, 2018: 1982–1991.
SUN B C, SAENKO K. Deep CORAL: correlation alignment for deep domain adaptation [M]//Computer vision — ECCV 2016 Workshops. Cham: Springer, 2016: 443–450.
GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al. Generative adversarial networks [J]. Communications of the ACM, 2020, 63(11): 139–144.
GANIN Y, LEMPITSKY V. Unsupervised domain adaptation by backpropagation [C]//32nd International Conference on Machine Learning. Lille, France: PMLR, 2015: 1180–1189.
GANIN Y, USTINOVA E, AJAKAN H, et al. Domain-adversarial training of neural networks [J]. Journal of Machine Learning Research, 2016, 17(1): 2096–2030.
WANG X M, LI L, YE W R, et al. Transferable attention for domain adaptation [J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33: 5345–5352.
MATSUURA T, HARADA T. Domain generalization using a mixture of multiple latent domains [J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7): 11749–11756.
WEI Y Y, ZHANG Z, WANG Y, et al. DerainCycle-GAN: Rain attentive CycleGAN for single image deraining and rainmaking [J]. IEEE Transactions on Image Processing, 2021, 30: 4788–4801.
GAO R, HOU X S, QIN J, et al. Zero-VAE-GAN: Generating unseen features for generalized and transductive zero-shot learning [J]. IEEE Transactions on Image Processing, 2020, 29: 3665–3680.
GAO X J, ZHANG Z, MU T T, et al. Self-attention driven adversarial similarity learning network [J]. Pattern Recognition, 2020, 105: 107331.
PEI Z, CAO Z, LONG M, et al. Multi-adversarial domain adaptation [J]//Proceedings of the AAAI Conference on Artificial Intelligence, 2018, 32(1): 3211–3218.
CHEN M H, ZHAO S, LIU H F, et al. Adversarial-learned loss for domain adaptation [J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(4): 3521–3528.
SAITO K, USHIKU Y, HARADA T. Asymmetric tri-training for unsupervised domain adaptation [C]//34th International Conference on Machine Learning. Sydney, Australia: PMLR, 2017: 2988–2997.
XIE S, ZHENG Z, CHEN L, et al. Learning semantic representations for unsupervised domain adaptation [C]//35th International Conference on Machine Learning. Stockholm, Sweden: PMLR, 2018: 5423–5432.
CHEN C Q, XIE W P, HUANG W B, et al. Progressive feature alignment for unsupervised domain adaptation [C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, CA, USA: IEEE, 2019: 627–636.
PAN Y W, YAO T, LI Y H, et al. Transferrable prototypical networks for unsupervised domain adaptation [C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, CA, USA: IEEE, 2019: 2234–2242.
ZOU Y, YU Z D, VIJAYA KUMAR B V K, et al. Unsupervised domain adaptation for semantic segmentation via class-balanced self-training [M]//Computer vision — ECCV 2018. Cham: Springer, 2018: 297–313.
WANG Q, BRECKON T. Unsupervised domain adaptation via structured prediction based selective pseudolabeling [J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(4): 6243–6250.
WANG M, DENG W H. Deep visual domain adaptation: A survey [J]. Neurocomputing, 2018, 312: 135–153.
HINTON G, VINYALS O, DEAN J. Distilling the Knowledge in a Neural Network [DB/OL]. (2015-05-09). https://arxiv.org/abs/1503.02531.
CHENG X, RAO Z F, CHEN Y L, et al. Explaining knowledge distillation by quantifying the knowledge [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, WA, USA: IEEE, 2020: 12922–12932.
YUAN L, TAY F E, LI G L, et al. Revisiting knowledge distillation via label smoothing regularization [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle, WA, USA: IEEE, 2020: 3902–3910.
SAENKO K, KULIS B, FRITZ M, et al. Adapting visual category models to new domains [M]//Computer vision — ECCV 2010. Berlin, Heidelberg: Springer, 2010: 213–226.
VENKATESWARA H, EUSEBIO J, CHAKRABORTY S, et al. Deep hashing network for unsupervised domain adaptation [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI, USA: IEEE, 2017: 5385–5394.
ZHU Y C, ZHUANG F Z, WANG J D, et al. Multi-representation adaptation network for cross-domain image classification [J]. Neural Networks, 2019, 119: 214–221.
LONG M, ZHU H, WANG J, et al. Deep transfer learning with joint adaptation networks [C]//34th International Conference on Machine Learning. Sydney, Australia: PMLR, 2017: 2208–2217.
BORGWARDT K M, GRETTON A, RASCH M J, et al. Integrating structured biological data by Kernel Maximum Mean Discrepancy [J]. Bioinformatics, 2006, 22(14): e49–e57.
ZELLINGER W, GRUBINGER T, LUGHOFER E, et al. Central moment discrepancy (cmd) for domain-invariant representation learning [C]//International Conference on Learning Representations. Toulon, France: Universite de Montreal, 2017: 234–245.
CHEN Q C, LIU Y, WANG Z W, et al. Re-weighted adversarial adaptation network for unsupervised domain adaptation [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA: IEEE, 2018: 7976–7985.
SANKARANARAYANAN S, BALAJI Y, CASTILLO C D, et al. Generate to adapt: Aligning domains using generative adversarial networks [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA: IEEE, 2018: 8503–8512.
VOLPI R, MORERIO P, SAVARESE S, et al. Adversarial feature augmentation for unsupervised domain adaptation [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA: IEEE, 2018: 5495–5504.
TZENG E, HOFFMAN J, SAENKO K, et al. Adversarial discriminative domain adaptation [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI, USA: IEEE, 2017: 2962–2971.
LIU H, LONG M, WANG J, et al. Transferable adversarial training: A general approach to adapting deep classifiers [C]//36th International Conference on Machine Learning. Long Beach, CA, USA: PMLR, 2019: 4013–4022.
SAITO K, WATANABE K, USHIKU Y, et al. Maximum classifier discrepancy for unsupervised domain adaptation [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA: IEEE, 2018: 3723–3732.
LU Z H, YANG Y X, ZHU X T, et al. Stochastic classifiers for unsupervised domain adaptation [C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Seattle, WA, USA: IEEE, 2020: 9108–9117.
HOFFMAN J, TZENG E, PARK T, et al. Cy-CADA: Cycle-consistent adversarial domain adaptation [C]//35th International Conference on Machine Learning. Stockholm, Sweden: PMLR, 2018: 1989–1998.
RUSSO P, CARLUCCI F M, TOMMASI T, et al. From source to target and back: Symmetric Bidirectional adaptive GAN [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City, UT, USA: IEEE, 2018: 8099–8108.
BOUSMALIS K, SILBERMAN N, DOHAN D, et al. Unsupervised pixel-level domain adaptation with generative adversarial networks [C]//2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI, USA: IEEE, 2017: 95–104.
LIU M Y, TUZEL O. Coupled generative adversarial networks [M]//Advances in Neural Information Processing Systems 29. Red Hook: Curran Associates Inc., 2016: 469–477.
KUMAR A, SATTIGERI P, WADHAWAN K, et al. Co-regularized alignment for unsupervised domain adaptation [C]//Advances in Neural Information Processing Systems 31. Red Hook: Curran Associates Inc., 2018: 543–555.
ZHANG Y, DAVID P, GONG B Q. Curriculum domain adaptation for semantic segmentation of urban scenes [C]//2017 IEEE International Conference on Computer Vision. Venice, Italy: IEEE, 2017: 2039–2049.
CHOI J, JEONG M, KIM T, et al Pseudo-labeling curriculum for unsupervised domain adaptation [DB/OL]. (2019-08-01). https://arxiv.org/abs/1908.00262.
HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas, NV, USA: IEEE, 2016: 770–778.
CHEN X, WANG S, LONG M, et al. Transferability vs. discriminability: Batch spectral penalization for adversarial domain adaptation [C]//36th International Conference on Machine Learning. Long Beach, CA, USA: PMLR, 2019: 1081–1090.
DENG J, DONG W, SOCHER R, et al. ImageNet: A large-scale hierarchical image database [C]//2009 IEEE Conference on Computer Vision and Pattern Recognition. Miami, FL, USA: IEEE, 2009: 248–255.
VAN DER MAATEN L, HINTON G. Visualizing data using t-SNE [J]. Journal of Machine Learning Research, 2008, 9(11): 2579–2605.
WU S, ZHONG J, CAO W M, et al. Improving domain-specific classification by collaborative learning with adaptation networks [J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2019, 33: 5450–5457.
SUN S L, CAO Z H, ZHU H, et al. A survey of optimization methods from a machine learning perspective [J]. IEEE Transactions on Cybernetics, 2020, 50(8): 3668–3681.
DAUPHIN Y, PASCANU R, GULCEHRE C, et al. Identifying and attacking the saddle point problem in high-dimensional non-convex optimization [M]//Advances in Neural Information Processing Systems 27. Red Hook: Curran Associates Inc., 2014: 2933–2941.
Author information
Authors and Affiliations
Corresponding author
Additional information
Foundation item: the 111 Project (No. BP0719010), and the Project of the Science and Technology Commission of Shanghai Municipality (No. 18DZ2270700)
Rights and permissions
About this article
Cite this article
Zhang, S., Lin, T. & Xu, Y. Boosting Unsupervised Domain Adaptation with Soft Pseudo-Label and Curriculum Learning. J. Shanghai Jiaotong Univ. (Sci.) 28, 703–716 (2023). https://doi.org/10.1007/s12204-022-2487-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12204-022-2487-5