Research map / Label-efficient and robust learning / Domain adaptation and generalization
Regularized Learning for Domain Adaptation under Label Shifts
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use RLLS importance weighting instead of BBSL | CIFAR-10, Minority-Class source shifts · CIFAR-10, Minority-Class source shifts | accuracy | Helped |
| Use RLLS importance weighting instead of BBSL or unweighted training | large target shifts · CIFAR-10, Dirichlet target shift | F-1 score: at most 20% higher | Helped |
| Partially apply regularized importance weights | RLLS in the low target-sample regime · CIFAR-10 and MNIST, low target sample sizes | accuracy: at least 10% | Helped |
| Use RLLS instead of BBSL | MNIST and CIFAR-10 across tested shift and sample-size settings · MNIST and CIFAR-10 | accuracy | Helped |
| Set the weight interpolation coefficient λ to an intermediate value (ablation) | RLLS · MNIST, Minority-Class source shift, low target sample sizes | accuracy | Helped |
| Set λ to 0 rather than applying estimated weights (ablation) | RLLS · MNIST, very few target samples | accuracy | Helped |
| Set λ to 1 and fully apply the estimated weights (ablation) | RLLS · MNIST, target sample size above 1000 | accuracy | Helped |
| Use a more corrupted black-box predictor for weight estimation (ablation) | RLLS · MNIST, varying target sample sizes | accuracy | Hurt |
| Apply importance weighting under a large Dirichlet source shift (ablation) | CIFAR-10, weighted methods including true and estimated weights · CIFAR-10, Dirichlet source shift and uniform target | accuracy | No clear effect |
| Use RLLS rather than BBSL for estimated importance weights | CIFAR-10 · CIFAR-10, source label shifts | weight-estimation MSE | Helped |
| Increase the number of minority classes in the source shift | RLLS and BBSL weight estimators · CIFAR-10, Minority-Class source shift | weight-estimation MSE | Helped |
| Use fully estimated weights instead of unweighted training (ablation) | RLLS · MNIST, minority-class source shift with p=0.01 | accuracy | Hurt |
Closest papers by meaning
- RLSbench: Domain Adaptation Under Relaxed Label Shift (ICML 2023 · similarity 0.92)
- Pseudo-Labeling for Unsupervised Domain Adaptation with Kernel GLMs (NeurIPS 2026 · similarity 0.88)
- Confusing Pair Correction Based on Category Prototype for Domain Adaptation under Noisy Environments (AAAI 2024 · similarity 0.87)
- Gradient-Based Sample Selection for Black-Box Universal Domain Adaptation (AAAI 2025 · similarity 0.87)
- A Theory of Learning Unified Model via Knowledge Integration from Label Space Varying Domains (CVPR 2025 · similarity 0.87)
- Label Shift Meets Online Learning: Ensuring Consistent Adaptation with Universal Dynamic Regret (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.