Research map / Label-efficient and robust learning / Domain adaptation and generalization
FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation
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 |
|---|---|---|---|
| add fixed-ratio mixup (ablation) | DANN · Office-31 | average accuracy: 82.2 -> 85.5 | Helped |
| add bidirectional matching (ablation) | DANN with fixed-ratio mixup · Office-31 | average accuracy: 85.5 -> 87.0 | Helped |
| add self-penalization (ablation) | DANN with fixed-ratio mixup and bidirectional matching · Office-31 | average accuracy: 87.0 -> 88.6 | Helped |
| add consistency regularization (ablation) | DANN with fixed-ratio mixup, bidirectional matching, and self-penalization · Office-31 | average accuracy: 88.6 -> 89.3 | Helped |
| replace fixed ratios with randomly sampled mixup ratios (ablation) | FixBi mixup and bidirectional matching · Office-31 A→W | accuracy: 89.3±0.4 -> 86.7±0.8 | Hurt |
| replace fixed ratios with range-limited random ratios (ablation) | FixBi mixup and bidirectional matching · Office-31 A→W | accuracy: 89.3±0.4 -> 83.3±6.2 | Hurt |
| add bidirectional matching to fixed-ratio mixup (ablation) | FixBi · Office-31 A→W | accuracy: 86.3±0.6 -> 89.3±0.4 | Helped |
| use the same mixup ratio for both models instead of complementary ratios (ablation) | FixBi ensemble · Office-31 A→W | accuracy: 90.1 vs 88.6; 90.1 vs 89.2 | Hurt |
| use two complementary fixed-ratio models rather than same-ratio models (ablation) | FixBi ensemble · Office-31 | average accuracy: 84.2 -> 87.0; 84.7 -> 87.0 | Helped |
| use FixBi rather than the strongest listed baseline | ResNet-50 UDA · Office-31 | average accuracy: 91.1 -> 91.4 | Helped |
| use FixBi rather than the strongest listed baseline | ResNet-50 UDA · Office-Home | average accuracy: 71.3 -> 72.7 | Helped |
| use FixBi rather than CAN | ResNet-101 UDA · VisDA-2017 | average accuracy: 87.2 vs 87.2 | No clear effect |
Closest papers by meaning
- Make the U in UDA Matter: Invariant Consistency Learning for Unsupervised Domain Adaptation (NeurIPS 2023 · similarity 0.93)
- Domain Adaptation with Invariant Representation Learning: What Transformations to Learn? (NeurIPS 2021 · similarity 0.91)
- On Learning Invariant Representations for Domain Adaptation (ICML 2019 · similarity 0.90)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.90)
- SimProF: A Simple Probabilistic Framework for Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.89)
- Tackling Dimensional Collapse toward Comprehensive Universal Domain Adaptation (ICML 2025 · similarity 0.89)
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.