Abstract
In recent years, deep neural networks have emerged as a dominant machine learning tool for a wide variety of application fields. Due to the expensive cost of manual labeling efforts, it is important to transfer knowledge from a label-rich source domain to an unlabeled target domain. The core problem is how to learn a domain-invariant representation to address the domain shift challenge, in which the training and test samples come from different distributions. First, considering the geometry of space probability distributions, we introduce an effective Hellinger Distance to match the source and target distributions on statistical manifold. Second, the data samples are not isolated individuals, and they are interrelated. The correlation information of data samples should not be neglected for domain adaptation. Distinguished from previous works, we pay attention to the correlation distributions over data samples. We design elaborately a Residual Graph Convolutional Network to construct the Instance Correlation Graph (ICG). The correlation information of data samples is exploited to reduce the domain shift. Therefore, a novel Instance Correlation Graph for Unsupervised Domain Adaptation is proposed, which is trained end-to-end by jointly optimizing three types of losses, i.e., Supervised Classification loss for source domain, Centroid Alignment loss to measure the centroid difference between source and target domain, ICG Alignment loss to match Instance Correlation Graph over two related domains. Extensive experiments are conducted on several hard transfer tasks to learn domain-invariant representations on three benchmarks: Office-31, Office-Home, and VisDA2017. Compared with other state-of-the-art techniques, our method achieves superior performance.
- [1] . 2014. Domain-adversarial neural networks. CoRR abs/1412.4446 (2014).Google Scholar
- [2] . 2002. Semi supervised logistic regression. In ECAI, (Ed.). 390–394. Google Scholar
Digital Library
- [3] . 2016. Concrete problems in AI safety. CoRR abs/1606.06565 (2016).Google Scholar
- [4] . 2019. Invariant risk minimization. CoRR abs/1907.02893 (2019).Google Scholar
- [5] . 2016. Layer normalization. CoRR abs/1607.06450 (2016).Google Scholar
- [6] . 2016. Distribution-matching embedding for visual domain adaptation. J. Mach. Learn. Res. 17 (2016), 108:1–108:30. Google Scholar
Digital Library
- [7] . 2010. A theory of learning from different domains. Mach. Learn. 79, 1–2 (2010), 151–175. Google Scholar
Digital Library
- [8] . 2006. Analysis of representations for domain adaptation. In NIPS. Google Scholar
Digital Library
- [9] . 2020. Robust domain adaptation: Representations, weights and inductive bias. CoRR abs/2006.13629.Google Scholar
- [10] . 2018. Transfer adversarial hashing for hamming space retrieval. In AAAI. Google Scholar
Digital Library
- [11] . 2009. Dimensionality reduction on statistical manifolds. Dissertations & Theses - Gradworks.Google Scholar
- [12] . 2019. Transferability vs. discriminability: Batch spectral penalization for adversarial domain adaptation. In ICML, Vol. 97. 1081–1090.Google Scholar
- [13] . 2020. Towards discriminability and diversity: Batch nuclear-norm maximization under label insufficient situations. CoRR abs/2003.12237.Google Scholar
- [14] . 2020. Gradually vanishing bridge for adversarial domain adaptation. In CVPR. 12452–12461.Google Scholar
- [15] . 2018. Unsupervised domain adaptation using regularized hyper-graph matching. In ICIP. 3758–3762.Google Scholar
- [16] . 2009. ImageNet: A large-scale hierarchical image database. In CVPR. 248–255.Google Scholar
- [17] . 2018. Graph adaptive knowledge transfer for unsupervised domain adaptation. In ECCV.Google Scholar
- [18] . 2015. Unsupervised domain adaptation by backpropagation. In ICML. 1180–1189. Google Scholar
Digital Library
- [19] . 2016. Domain-adversarial training of neural networks. J. Mach. Learn. Res. 17 (2016), 59:1–59:35. Google Scholar
Digital Library
- [20] . 2017. Fine-grained recognition in the wild: A multi-task domain adaptation approach. In ICCV.Google Scholar
- [21] . 2016. Deep reconstruction-classification networks for unsupervised domain adaptation. In ECCV. 597–613.Google Scholar
- [22] . 2012. Geodesic flow kernel for unsupervised domain adaptation. In CVPR.Google Scholar
- [23] . 2019. DLOW: Domain flow for adaptation and generalization. In CVPR. 2477–2486.Google Scholar
- [24] . 2014. Generative adversarial nets. In NIPS. Google Scholar
Digital Library
- [25] . 2017. Accurate, large minibatch SGD: Training imagenet in 1 hour. CoRR abs/1706.02677.Google Scholar
- [26] . 2012. A kernel two-sample test. J. Mach. Learn. Res. 13, 1 (2012), 723–773. Google Scholar
Digital Library
- [27] . 2012. Optimal kernel choice for large-scale two-sample tests. In NIPS. Google Scholar
Digital Library
- [28] . 2013. Rotation averaging. Int. J. Comput. Vis. 103, 3 (2013), 267–305.Google Scholar
Cross Ref
- [29] . 2016. Deep residual learning for image recognition. In CVPR. 770–778.Google Scholar
- [30] . 2018. CyCADA: Cycle-consistent adversarial domain adaptation. In ICML, Vol. 80. 1994–2003.Google Scholar
- [31] . 2018. Multimodal unsupervised image-to-image translation. In ECCV, Vol. 11207. 179–196.Google Scholar
- [32] . 2019. Contrastive adaptation network for unsupervised domain adaptation. In CVPR. 4893–4902.Google Scholar
- [33] . 2011. Geometrical Foundations of Asymptotic Inference. Vol. 908.Google Scholar
- [34] . 2017. Semi-supervised classification with graph convolutional networks. In ICLR.Google Scholar
- [35] . 2013. 3D object representations for fine-grained categorization. In ICCV Workshops. 554–561. Google Scholar
Digital Library
- [36] . 2012. ImageNet classification with deep convolutional neural networks. In NIPS. 1106–1114. Google Scholar
Digital Library
- [37] . 1998. Gradient-based learning applied to document recognition. Proc. IEEE 86, 11 (1998), 2278–2324.Google Scholar
Cross Ref
- [38] . 2019. Locality preserving joint transfer for domain adaptation. IEEE Trans. Image Process. 28, 12 (2019), 6103–6115.Google Scholar
- [39] . 2017. Feature pyramid networks for object detection. In CVPR. 936–944.Google Scholar
- [40] . 2019. Feature-level frankenstein: Eliminating variations for discriminative recognition. In CVPR. 637–646.Google Scholar
- [41] . 2015. Learning transferable features with deep adaptation networks. In ICML. Google Scholar
Digital Library
- [42] . 2018. Conditional adversarial domain adaptation. In NeurIPS. 1647–1657. Google Scholar
Digital Library
- [43] . 2013. Transfer feature learning with joint distribution adaptation. In ICCV. Google Scholar
Digital Library
- [44] . 2016. Unsupervised domain adaptation with residual transfer networks. In NIPS. Google Scholar
Digital Library
- [45] . 2017. Deep transfer learning with joint adaptation networks. In ICML. Google Scholar
Digital Library
- [46] . 2017. Label efficient learning of transferable representations acrosss domains and tasks. In NIPS. 165–177. Google Scholar
Digital Library
- [47] . 2019. GCAN: Graph convolutional adversarial network for unsupervised domain adaptation. In CVPR.Google Scholar
- [48] . 2011. Domain adaptation via transfer component analysis. IEEE Trans. Neur. Netw. 22, 2 (2011), 199–210. Google Scholar
Digital Library
- [49] . 2018. Multi-adversarial domain adaptation. In AAAI. Google Scholar
Digital Library
- [50] . 2017. Visda: The visual domain adaptation challenge. CoRR abs/1710.06924.Google Scholar
- [51] . 2020. Domain adaptation on graphs by learning aligned graph bases. IEEE Trans. Knowl. Data Eng. (2020), 1–1.Google Scholar
- [52] . 2018. Unsupervised domain adaptation with similarity learning. In CVPR. 8004–8013.Google Scholar
- [53] . 2019. Fine-tuning CNN image retrieval with no human annotation. IEEE Trans. Pattern Anal. Mach. Intell. 41, 7 (2019), 1655–1668.Google Scholar
Cross Ref
- [54] . 2010. Adapting visual category models to new domains. In ECCV. Google Scholar
Digital Library
- [55] . 2018. Adversarial dropout regularization. In ICLR.Google Scholar
- [56] . 2018. Generate to adapt: Aligning domains using generative adversarial networks. In CVPR. 8503–8512.Google Scholar
- [57] . 2017. Fully convolutional networks for semantic segmentation. IEEE Trans. Pattern Anal. Mach. Intell. 39, 4 (2017), 640–651. Google Scholar
Digital Library
- [58] . 2017. Prototypical networks for few-shot learning. In NIPS. 4077–4087. Google Scholar
Digital Library
- [59] . 2016. Return of frustratingly easy domain adaptation. In AAAI. Google Scholar
Digital Library
- [60] . 2016. Deep CORAL: Correlation alignment for deep domain adaptation. In ECCV Workshops.Google Scholar
- [61] . 2019. Towards high-fidelity nonlinear 3D face morphable model. In CVPR.Google Scholar
- [62] . 2018. Learning to adapt structured output space for semantic segmentation. In CVPR. 7472–7481.Google Scholar
- [63] . 2015. Simultaneous deep transfer across domains and tasks. In ICCV. Google Scholar
Digital Library
- [64] . 2017. Adversarial discriminative domain adaptation. In CVPR.Google Scholar
- [65] . 2014. Deep domain confusion: Maximizing for domain invariance. CoRR abs/1412.3474 (2014).Google Scholar
- [66] . 2017. Attention is all you need. In NIPS. 5998–6008. Google Scholar
Digital Library
- [67] . 2017. Deep hashing network for unsupervised domain adaptation. In CVPR.Google Scholar
- [68] . 2018. Non-local neural networks. In CVPR.Google Scholar
- [69] . 2018. Videos as space-time region graphs. In ECCV. 413–431.Google Scholar
- [70] . 2018. Learning semantic representations for unsupervised domain adaptation. In ICML.Google Scholar
- [71] . 2020. Adversarial domain adaptation with domain mixup. In AAAI. 6502–6509.Google Scholar
- [72] . 2019. Larger norm more transferable: An adaptive feature norm approach for unsupervised domain adaptation. In ICCV. 1426–1435.Google Scholar
- [73] . 2017. Mind the class weight bias: Weighted maximum mean discrepancy for unsupervised domain adaptation. In CVPR.Google Scholar
- [74] . 2019. Cross-domain visual representations via unsupervised graph alignment. In AAAI. 5613–5620. Google Scholar
Digital Library
- [75] . 2017. Central moment discrepancy (CMD) for domain-invariant representation learning. In ICLR.Google Scholar
- [76] . 2021. Mixture distribution graph network for few shot learning. IEEE Trans. Cogn. Dev. Syst. (2021), 1–1. https://doi.org/10.1109/TCDS.2021.3075280Google Scholar
- [77] . 2019. Bridging theory and algorithm for domain adaptation. In ICML. 7404–7413.Google Scholar
- [78] . 2019. Domain-symmetric networks for adversarial domain adaptation. In CVPR. 5031–5040.Google Scholar
- [79] . 2010. Cross validation framework to choose amongst models and datasets for transfer learning. In ECML. Google Scholar
Digital Library
- [80] . 2017. Unpaired image-to-image translation using cycle-consistent adversarial networks. In ICCV. 2242–2251.Google Scholar
- [81] . 2017. Deep unsupervised convolutional domain adaptation. In ACM MM. 261–269. Google Scholar
Digital Library
Index Terms
Instance Correlation Graph for Unsupervised Domain Adaptation
Recommendations
A novel class restriction loss for unsupervised domain adaptation
AbstractDomain adaptation has demonstrated promising performance to learn models in new environments. It aims to transfer the knowledge learned in labeled source domain to a new target domain, avoiding expensive efforts of label annotation. ...
Soft Labels Transfer with Discriminative Representations Learning for Unsupervised Domain Adaptation
Machine Learning and Knowledge Discovery in DatabasesAbstractDomain adaptation aims to address the challenge of transferring the knowledge obtained from the source domain with rich label information to the target domain with less or even no label information. Recent methods start to tackle this problem by ...
Self-corrected unsupervised domain adaptation
AbstractUnsupervised domain adaptation (UDA), which aims to use knowledge from a label-rich source domain to help learn unlabeled target domain, has recently attracted much attention. UDA methods mainly concentrate on source classification and ...






Comments