Visual Domain Adaptation with Manifold Embedded Distribution Alignment

@article{Wang2018VisualDA,
  title={Visual Domain Adaptation with Manifold Embedded Distribution Alignment},
  author={Jindong Wang and Wenjie Feng and Yiqiang Chen and Han Yu and Meiyu Huang and Philip S. Yu},
  journal={Proceedings of the 26th ACM international conference on Multimedia},
  year={2018}
}
Visual domain adaptation aims to learn robust classifiers for the target domain by leveraging knowledge from a source domain. Existing methods either attempt to align the cross-domain distributions, or perform manifold subspace learning. However, there are two significant challenges: (1) degenerated feature transformation, which means that distribution alignment is often performed in the original feature space, where feature distortions are hard to overcome. On the other hand, subspace learning… 

Figures and Tables from this paper

Bures Joint Distribution Alignment with Dynamic Margin for Unsupervised Domain Adaptation
TLDR
The Bures Joint Distribution Alignment (BJDA) algorithm is proposed which directly models the joint distribution shift based on the optimal transport theory in the infinitedimensional kernel spaces and introduces a dynamic margin in contrastive learning phase to flexibly characterize the class separability and improve the discriminative ability of representations.
A Subspace Based Transfer Joint Matching with Laplacian Regularization for Visual Domain Adaptation
TLDR
A novel Subspace based Transfer Joint Matching with Laplacian Regularization (STJML) method for visual domain adaptation by jointly matching the features and re-weighting the instances across different domains is proposed.
Canonical Correlation Cross-Domain Alignment for Unsupervised Domain Adaptation
TLDR
A new method, canonical correlation cross-domain alignment (CCCA), is proposed to effectively reduce the cross- domain distribution difference by combining the least squares formula of CCA with domain adaptation.
Joint distinct subspace learning and unsupervised transfer classification for visual domain adaptation
TLDR
For subspace alignment, weighted joint geometrical and statistical alignment is proposed to learn two coupled projections for mapping the source and target data into respective subspaces by accounting the importance of marginal and conditional distributions, differently.
Deep Spherical Manifold Gaussian Kernel for Unsupervised Domain Adaptation
  • Youshan Zhang, B. Davison
  • Computer Science
    2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)
  • 2021
TLDR
A deep spherical manifold Gaussian kernel (DSGK) framework is proposed to map the source and target subspaces into a spherical manifold and reduce the discrepancy between them by embedding both extracted features and a Gaussiankernel.
Joint discriminative subspace and distribution adaptation for unsupervised domain adaptation
TLDR
The experiment results show that DSL, GSDA and DSL-GSDA outperform other state-of-the-art domain adaptation methods by 6.19%, 1.48% and 1.99% improvement, respectively.
Domain Adaptation and Image Classification via Deep Conditional Adaptation Network
TLDR
This paper proposes a novel unsupervised domain adaptation method, Deep Conditional Adaptation Network (DCAN), based on conditional distribution alignment of feature spaces, which reduces the domain discrepancy by minimizing the Conditional Maximum Mean Discrepancy between the conditional distributions of deep features on the source and target domains.
Adaptive domain of dynamic distribution based on manifold space
  • Daoyuan Yu, Xuegang Hu
  • Computer Science
    2020 IEEE International Conference on Knowledge Graph (ICKG)
  • 2020
TLDR
The structural risk minimization and an improved dynamic manifold regularization constraint are proposed to solve the problems of degenerated feature transformation and unevaluated distribution alignment in the process of feature representation.
...
...

References

SHOWING 1-10 OF 47 REFERENCES
Distribution-Matching Embedding for Visual Domain Adaptation
TLDR
This paper introduces a Distribution-Matching Embedding approach: An unsupervised domain adaptation method that overcomes this issue by mapping the data to a latent space where the distance between the empirical distributions of the source and target examples is minimized.
Visual domain adaptation via transfer feature learning
TLDR
A novel transfer learning and domain adaptation approach, referred to as visual domain adaptation (VDA), which reduces the joint marginal and conditional distributions across domains in an unsupervised manner where no label is available in test set.
Unsupervised Domain Adaptation by Domain Invariant Projection
TLDR
This paper learns a projection of the data to a low-dimensional latent space where the distance between the empirical distributions of the source and target examples is minimized and demonstrates the effectiveness of the approach on the task of visual object recognition.
Subspace Distribution Alignment for Unsupervised Domain Adaptation
TLDR
A unified view of existing subspace mapping based methods is presented and a generalized approach that also aligns the distributions as well as the subspace bases is developed that shows improved results over published approaches.
Transfer Feature Learning with Joint Distribution Adaptation
TLDR
JDA aims to jointly adapt both the marginal distribution and conditional distribution in a principled dimensionality reduction procedure, and construct new feature representation that is effective and robust for substantial distribution difference.
Transfer Joint Matching for Unsupervised Domain Adaptation
TLDR
This paper aims to reduce the domain difference by jointly matching the features and reweighting the instances across domains in a principled dimensionality reduction procedure, and construct new feature representation that is invariant to both the distribution difference and the irrelevant instances.
Geodesic flow kernel for unsupervised domain adaptation
TLDR
This paper proposes a new kernel-based method that takes advantage of low-dimensional structures that are intrinsic to many vision datasets, and introduces a metric that reliably measures the adaptability between a pair of source and target domains.
Deep Unsupervised Convolutional Domain Adaptation
TLDR
Deep Unsupervised Convolutional Domain Adaptation DUCDA method is proposed, which jointly minimizes the supervised classification loss of labeled source data and the unsupervised correlation alignment loss measured on both convolutional layers and fully connected layers.
Domain Adaptation on the Statistical Manifold
TLDR
This framework introduces a sample selection method and a subspace-based method for unsupervised domain adaptation, and shows that both these manifold-based techniques outperform the corresponding approaches based on the MMD.
A Unified Framework for Metric Transfer Learning
TLDR
A metric transfer learning framework (MTF) is proposed to encode metric learning in transfer learning to make knowledge transfer across domains more effective and develops general solutions to both classification and regression problems on top of MTLF.
...
...