Research map / Label-efficient and robust learning / Domain adaptation and generalization
Spherical Space Domain Adaptation With Robust Pseudo-Label Loss
Check 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 the full spherical adaptation method with robust pseudo-label and conditional entropy losses instead of DANN | DANN · Office-31 | accuracy (%): 8.0% | Helped |
| Use the full spherical adaptation method instead of MSTN | MSTN · Office-31 | accuracy (%): 4.6% | Helped |
| Use the full spherical adaptation method instead of DANN | DANN · ImageCLEF-DA | accuracy (%): 5.1% | Helped |
| Use the full spherical adaptation method instead of DANN | DANN · Office-Home | accuracy (%): 12.2% | Helped |
| Add robust pseudo-label loss to spherical DANN (ablation) | DANN+S · Office-31, ImageCLEF-DA, Office-Home | accuracy (%): Table 5: 86.7 -> 89.2; 88.5 -> 89.4; 59.8 -> 68.4 | Helped |
| Add robust pseudo-label loss in Euclidean space (ablation) | DANN · Office-31, ImageCLEF-DA, Office-Home | accuracy (%): Table 5: 82.2 -> 88.7; 85.0 -> 89.1; 57.6 -> 67.3 | Helped |
| Add robust pseudo-label loss to spherical DANN rather than using it in Euclidean space (ablation) | DANN · Office-31, ImageCLEF-DA, Office-Home | accuracy (%): Table 5: 89.2 vs 88.7; 89.4 vs 89.1; 68.4 vs 67.3 | Mixed |
| Add conditional entropy loss to spherical DANN with robust pseudo-label loss (ablation) | DANN+S+R · Office-31, ImageCLEF-DA, Office-Home | accuracy (%): Table 5: 89.2 -> 90.2; 89.4 -> 90.1; 68.4 -> 69.8 | Helped |
| Use spherical features, classifier, and discriminator instead of Euclidean DANN (ablation) | DANN · Office-31, ImageCLEF-DA, Office-Home, VisDA-2017 | accuracy (%): 4.5%, 3.5%, 2.2%, 3.9% | Helped |
| Normalize features but keep the classifier and discriminator Euclidean (ablation) | DANN · Office-31 | accuracy (%): Table 6: 84.7 vs 82.2 | Helped |
| Replace the Euclidean classifier and discriminator with spherical versions for normalized features (ablation) | DANN+S (Eucl) · Office-31 | accuracy (%): 2.0% | Helped |
| Omit spherical exponential and logarithmic maps from spherical perceptron layers (ablation) | DANN+S · Office-31 | accuracy (%): Table 6: 86.1 vs 86.7 | Hurt |
Closest papers by meaning
- Adversarial-Learned Loss for Domain Adaptation (AAAI 2020 · similarity 0.85)
- Domain Adaptation with Conditional Distribution Matching and Generalized Label Shift (NeurIPS 2020 · similarity 0.85)
- Toward Improving Robustness and Accuracy in Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.82)
- On Unsupervised Domain Adaptation: Pseudo Label Guided Mixup for Adversarial Prompt Tuning (AAAI 2024 · similarity 0.82)
- Pseudo-Labeling for Unsupervised Domain Adaptation with Kernel GLMs (NeurIPS 2026 · similarity 0.82)
- CASUAL: Conditional Support Alignment for Domain Adaptation with Label Shift (AAAI 2025 · similarity 0.82)
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.