mlsanityData to 7 Oct 2026

Equivariant Disentangled Transformation for Domain Generalization under Combination Shift

arXiv 2208.02011 ↗First posted 3 Aug 2022cs.LGcited by 1

Not on the map. Why?

Abstract (arXiv)

Machine learning systems may encounter unexpected problems when the data distribution changes in the deployment environment. A major reason is that certain combinations of domains and labels are not observed during training but appear in the test environment. Although various invariance-based algorithms can be applied, we find that the performance gain is often marginal. To formally analyze this issue, we provide a unique algebraic formulation of the combination shift problem based on the concepts of homomorphism, equivariance, and a refined definition of disentanglement. The algebraic requirements naturally derive a simple yet effective method, referred to as equivariant disentangled transformation (EDT), which augments the data based on the algebraic structures of labels and makes the transformation satisfy the equivariance and disentanglement requirements. Experimental results demonstrate that invariance may be insufficient, and it is important to exploit the equivariance structure in the combination shift problem.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 43 of its 79 references that are in the map. First posted on arXiv 3 Aug 2022.

Builds on

43 of 79 references.

References · 43

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  3. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,155
  4. 2013Auto-Encoding Variational Bayescited by 3,980
  5. 2015Domain-Adversarial Training of Neural Networkscited by 3,023
  6. 2017Improved Training of Wasserstein GANscited by 2,223
  7. 2019Invariant Risk Minimizationcited by 1,832
  8. 2018Spectral Normalization for Generative Adversarial Networkscited by 1,618
  9. 2016Deep CORAL: Correlation Alignment for Deep Domain Adaptationcited by 1,194
  10. 2020In Search of Lost Domain Generalizationcited by 1,042
  11. 2017StarGAN: Unified Generative Adversarial Networks for Multi-Domain Image-to-Image Translationcited by 1,036
  12. 2017Deeper, Broader and Artier Domain Generalizationcited by 983
31 more
  1. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 929
  2. 2021Generalizing to Unseen Domains: A Survey on Domain Generalizationcited by 762
  3. 2020Learning from Noisy Labels with Deep Neural Networks: A Surveycited by 580
  4. 2021Domain Generalization with MixStylecited by 563
  5. 2018Generalizing to Unseen Domains via Adversarial Data Augmentationcited by 451
  6. 2021Deep Long-Tailed Learning: A Surveycited by 416
  7. 2016Kernel Mean Embedding of Distributions: A Review and Beyondcited by 360
  8. 2020Environment Inference for Invariant Learningcited by 306
  9. 2018Towards a Definition of Disentangled Representationscited by 294
  10. 2021Gradient Matching for Domain Generalizationcited by 258
  11. 2015Generative Moment Matching Networkscited by 252
  12. 2019Causality matters in medical imagingcited by 193
  13. 2021A Fine-Grained Analysis on Distribution Shiftcited by 171
  14. 2015Training generative neural networks via Maximum Mean Discrepancy optimizationcited by 149
  15. 2018Strike (with) a Pose: Neural Networks Are Easily Fooled by Strange Poses of Familiar Objectscited by 143
  16. 2019Generalizing to unseen domains via distribution matchingcited by 109
  17. 2018Robustly Disentangled Causal Mechanisms: Validating Deep Representations for Interventional Robustnesscited by 108
  18. 2020On Disentangled Representations Learned From Correlated Datacited by 94
  19. 2017Learning to Compose Domain-Specific Transformations for Data Augmentationcited by 91
  20. 2019Weakly Supervised Disentanglement with Guaranteescited by 82
  21. 2020Model Patching: Closing the Subgroup Performance Gap with Data Augmentationcited by 82
  22. 2020Continuously Indexed Domain Adaptationcited by 74
  23. 2020Regularizing Deep Networks with Semantic Data Augmentationcited by 73
  24. 2021Visual Representation Learning Does Not Generalize Strongly Within the Same Domaincited by 60
  25. 2018Leveraging Unlabeled Data for Crowd Counting by Learning to Rankcited by 57
  26. 2021Self-Supervised Learning Disentangled Group Representation as Featurecited by 47
  27. 2020Debiasing Skin Lesion Datasets and Models? Not So Fastcited by 43
  28. 2019Symmetry-Based Disentangled Representation Learning requires Interaction with Environmentscited by 35
  29. 2021Towards Building A Group-based Unsupervised Representation Disentanglement Frameworkcited by 22
  30. 2020Linear Disentangled Representations and Unsupervised Action Estimationcited by 14
  31. 2021Disentanglement Analysis with Partial Information Decompositioncited by 9

Built on by

Cited by year

Citing papers per year: 2023: 1.

1 paper.

  1. 2023Correcting Flaws in Common Disentanglement Metricscited by 2