Research map / Label-efficient and robust learning / Domain adaptation and generalization
Adversarial Reweighting for Partial Domain Adaptation
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 |
|---|---|---|---|
| add adversarial reweighting of source data (ablation) | Cls+Ent · VisDA-2017 PDA | accuracy (%): 75.68 -> 83.62 | Helped |
| add adversarial reweighting of source data (ablation) | Cls+Ent · DomainNet PDA | accuracy (%): 64.06 -> 65.47 | Helped |
| add adversarial reweighting of source data (ablation) | Cls+Ent · Office-Home PDA | accuracy (%): 74.68 -> 78.29 | Helped |
| add adversarial reweighting of source data (ablation) | Cls+Ent · ImageNet-Caltech PDA | accuracy (%): 81.10 -> 84.69 | Helped |
| add adversarial reweighting of source data (ablation) | Cls+Ent · Office-31 PDA | accuracy (%): 91.66 -> 96.93 | Helped |
| add AR on top of AUS loss (ablation) | Cls+Ent+AUS · Office-Home PDA | accuracy (%): 76.08 -> 79.43 | Helped |
| add AR on top of label smoothing (ablation) | Cls+Ent+LS · Office-Home PDA | accuracy (%): 76.80 -> 79.44 | Helped |
| use linear classifier instead of spherical logistic regression in AR (ablation) | AR · VisDA-2017 PDA | accuracy (%): 83.62 -> 80.09 | Hurt |
| use classifier-based or discriminator-based reweighting instead of adversarial reweighting (ablation) | AR · VisDA-2017 / Office-Home | accuracy (%) | Hurt |
| learn source weights with JS divergence or MMD instead of Wasserstein distance (ablation) | AR · Office-Home | accuracy (%) | Hurt |
| learn one weight per class instead of one per sample (ablation) | AR · Office-Home | accuracy (%) | Hurt |
| add reweighted feature distribution alignment losses (PADA, BAA, MMD) | baseline w/o Align · VisDA-2017 and DomainNet | accuracy (%) | Hurt |
Closest papers by meaning
- Importance Weighted Adversarial Nets for Partial Domain Adaptation (CVPR 2018 · similarity 0.91)
- Re-Weighted Adversarial Adaptation Network for Unsupervised Domain Adaptation (CVPR 2018 · similarity 0.91)
- Adversarial Discriminative Domain Adaptation (CVPR 2017 · similarity 0.89)
- Learning to Transfer Examples for Partial Domain Adaptation (CVPR 2019 · similarity 0.89)
- Improving Accuracy-robustness Trade-off via Pixel Reweighted Adversarial Training (ICML 2024 · similarity 0.86)
- Theoretical Performance Guarantees for Partial Domain Adaptation via Partial Optimal Transport (ICML 2025 · similarity 0.86)
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.