Research map / Label-efficient and robust learning / Domain adaptation and generalization
Structure-Aware Feature Fusion for Unsupervised 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 global MI loss (global feature vs classifier prediction) (ablation) | DANN base · Office-31 A->W | accuracy (%): 82.0 -> 94.2 | Helped |
| add local MI loss (global vs local feature) (ablation) | DANN base · Office-31 A->W | accuracy (%): 82.0 -> 92.2 | Helped |
| add both MI losses to DANN (ablation) | DANN base · Office-31 A->W | accuracy (%): 82.0 -> 96.4 | Helped |
| add global MI loss (ablation) | MMD base · Office-31 A->W | accuracy (%): 80.5 -> 88.0 | Helped |
| add local MI loss (ablation) | MMD base · Office-31 A->W | accuracy (%): 80.5 -> 88.3 | Helped |
| add both MI losses to MMD (ablation) | MMD base · Office-31 A->W | accuracy (%): 80.5 -> 90.2 | Helped |
| MI-based integration instead of concatenation after projection (ablation) | STAFF · Office-31 A->W | accuracy (%): 89.1 -> 96.4 | Helped |
| MI-based integration instead of multiple adversarial discriminators (ablation) | STAFF · Office-31 A->W | accuracy (%): 83.5 -> 96.4 | Helped |
| MI-based integration lowers A-distance versus concatenation (ablation) | STAFF · Office-31 A->W | A-distance: 1.451 -> 1.313 | Helped |
| use NCE instead of JSD for MI estimation (batch size 32) (ablation) | STAFF · Office-31 A->W | accuracy (%): 96.4 -> 91.0 | Hurt |
| use MINE instead of JSD for MI estimation (ablation) | STAFF · Office-31 A->W | accuracy (%): 96.4 -> 93.7 | Hurt |
| increase batch size to 256 for NCE loss (ablation) | STAFF with NCE · Office-31 A->W | accuracy (%): 91.0 -> 96.3 | Helped |
Closest papers by meaning
- Progressive Feature Alignment for Unsupervised Domain Adaptation (CVPR 2019 · similarity 0.91)
- Exploiting Local Feature Patterns for Unsupervised Domain Adaptation (AAAI 2019 · similarity 0.91)
- Universal Domain Adaptation for Semantic Segmentation (CVPR 2025 · similarity 0.88)
- SimProF: A Simple Probabilistic Framework for Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.88)
- Enhancing Domain Adaptation through Prompt Gradient Alignment (NeurIPS 2024 · similarity 0.88)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.88)
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.