Research map / Label-efficient and robust learning / Domain adaptation and generalization
Unknown-Aware Domain Adversarial Learning for Open-Set Domain Adaptation
JoonHo Jang, Byeonghu Na, Dong Hyeok Shin, Mingi Ji, Kyungwoo Song, Il-chul Moon
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| replace STA's known-only feature alignment with unknown-aware alignment (keeping STA weights) (ablation) | STA · Office-31 (ResNet-50) | HOS avg: 72.5 -> 85.5 | Improved |
| replace ROS alignment with unknown-aware alignment (keeping ROS weights) (ablation) | ROS · Office-31 (ResNet-50) | HOS avg: 85.9 -> 86.7 | Improved |
| use UADAL's own posterior-inference weights instead of STA/ROS weights with UADAL core (ablation) | UADAL · Office-31 (ResNet-50) | HOS avg: 86.7 -> 88.1 | Improved |
| add conditional discriminator input (cUADAL) over UADAL (ablation) | UADAL · VisDA | HOS: 69.4 -> 70.1 (ResNet-50) | Improved |
| add conditional discriminator input (cUADAL) over UADAL (ablation) | UADAL · Office-Home | HOS avg: 68.7 -> 68.5 (ResNet-50); 64.1 -> 64.2 (EfficientNet-B0) | Mixed |
| posterior-inference (Beta mixture) open-set recognition instead of thresholding normalized entropy (ablation) | UADAL · Office-31 | HOS | Improved |
| update posterior inference frequently instead of only initially (ablation) | UADAL · Office-31 | HOS | Improved |
| fit the mixture model on a 10% sample of target instances instead of all (ablation) | UADAL posterior inference · Office-31 and Office-Home | HOS | No clear effect |
| add entropy minimization loss (ablation) | UADAL · Office-31 and Office-Home | HOS | Improved |
Closest papers by meaning
- Multi-Source Open-Set Deep Adversarial Domain Adaptation (ECCV 2020 · similarity 0.91)
- Open Set Domain Adaptation by Backpropagation (ECCV 2018 · similarity 0.90)
- A Balanced and Uncertainty-aware Approach for Partial Domain Adaptation (ECCV 2020 · similarity 0.89)
- Unknown-Oriented Learning for Open Set Domain Adaptation (ECCV 2022 · similarity 0.89)
- A DIRT-T Approach to Unsupervised Domain Adaptation (ICLR 2018 · similarity 0.89)
- Co-regularized Alignment for Unsupervised Domain Adaptation (NeurIPS 2018 · 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.
