Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Decorrelation with Potential Energy Ranking
Sen Pei, Jiaxi Sun, Richard Yi Da Xu, Shiming Xiang, Gaofeng Meng
AI-extracted. The measured effects below were read by AI from the paper's full text (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 PoER ranking and cluster regularization (ablation) | ResNet-18 with GCPL classification · PACS | top-1 accuracy: 81.86 -> 84.18 | Improved |
| add shallow-layer ranking loss (ablation) | ResNet-18 with GCPL classification · PACS | top-1 accuracy: 81.86 -> 83.72 | Improved |
| add deeper-layer cluster loss (ablation) | ResNet-18 with GCPL classification · PACS | top-1 accuracy: 81.86 -> 82.60 | Improved |
| add ranking loss alongside cluster loss (ablation) | PoER with cluster loss only · PACS | top-1 accuracy: 82.60 -> 84.18 | Improved |
| set regularization weight α to 0.2 rather than 0.0 (ablation) | PoER · PACS | top-1 accuracy: 81.86 -> 84.20 | Improved |
| increase prototype count from 1 to 10 (ablation) | PoER · PACS | top-1 accuracy: 82.40 -> 84.21 | Improved |
| use PoER instead of GCPL | ResNet-18 · PACS Avg. | top-1 accuracy: 81.86 -> 84.18 | Improved |
| use PoER instead of DDAIG | ResNet-18 · Office-Home Avg. | top-1 accuracy: 65.5 -> 66.6 | Improved |
| use PoER instead of DDAIG | ResNet-18 · Digits-DG Avg. | top-1 accuracy: 77.6 -> 81.8 | Improved |
| use PoER instead of StableNet | domain generalization method · VLCS Avg. | top-1 accuracy: 77.65 -> 76.63 | Hurt |
| use PoER rather than the vanilla ResNet-18 (ablation) | ResNet-18 · NICO | top-1 accuracy: 51.71 -> 62.62 | Improved |
| add PoER to convolutional and attention-based backbones (ablation) | ResNet-50, DenseNet-121, and Swin-T · NICO | top-1 accuracy | Improved |
Closest papers by meaning
- Beyond Sharing Weights for Deep Domain Adaptation (TPAMI 2018 · similarity 0.85)
- Exploring Object Relation in Mean Teacher for Cross-Domain Detection (CVPR 2019 · similarity 0.85)
- Label-Focused Inductive Bias over Latent Object Features in Visual Classification (ICLR 2024 · similarity 0.85)
- Open Domain Generalization with Domain-Augmented Meta-Learning (CVPR 2021 · similarity 0.84)
- A Simple Feature Augmentation for Domain Generalization (ICCV 2021 · similarity 0.84)
- Semantic Feature Learning for Universal Unsupervised Cross-Domain Retrieval (NeurIPS 2024 · similarity 0.83)
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.
