Research map / Label-efficient and robust learning / Domain adaptation and generalization
A Closer Look at Smoothness in Domain Adversarial Training
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 |
|---|---|---|---|
| Apply SDAT to CDAN + MCC | CDAN + MCC with ResNet-50 · Office-Home | accuracy (Avg, %): 71.3 -> 72.2 | Helped |
| Apply SDAT to CDAN + MCC | CDAN + MCC with ViT-B/16 · Office-Home | accuracy (Avg, %): 82.2 -> 84.3 | Helped |
| Apply SDAT to CDAN | CDAN with ResNet-101 · VisDA-2017 | mean accuracy (%): 80.6 -> 82.1 | Helped |
| Apply SDAT to CDAN + MCC | CDAN + MCC with ViT-B/16 · VisDA-2017 | mean accuracy (%): 87.7 -> 89.8 | Helped |
| Apply SDAT to CDAN | CDAN with ResNet-101 · DomainNet | accuracy (Avg, %): 40.2 -> 42.1 | Helped |
| Smooth classification loss only instead of both classification and regression losses (ablation) | DA-Faster with ResNet-101 · Pascal to Clipart (bs=2) | mAP: 29.04 -> 31.23 | Helped |
| Smooth classification loss only instead of both classification and regression losses (ablation) | DA-Faster with ResNet-50 · Cityscapes to Foggy Cityscapes (bs=2) | mAP: 37.47 -> 38.00 | Helped |
| Smooth classification loss only instead of both classification and regression losses (ablation) | DA-Faster with ResNet-101 · Pascal to Clipart (bs=8) | mAP: 27.64 -> 30.74 | Helped |
| Smooth the adversarial loss (ablation) | DAT · Art to Clipart | accuracy (%): 51.0 | Hurt |
| Increase adversarial-loss smoothing strength (ablation) | DAT discriminator · Art to Clipart | target accuracy | Hurt |
| Use SAM for source-only ERM (ablation) | ERM with ViT-B/16 · VisDA-2017 source-only | mean accuracy (%): 69.5 -> 67.1 | Hurt |
| Use SAM for source-only ERM (ablation) | ERM with ViT-B/16 · Office-Home source-only | accuracy (Avg, %): 73.9 -> 72.8 | Hurt |
Closest papers by meaning
- Sharpness-Aware Model-Agnostic Long-Tailed Domain Generalization (AAAI 2024 · similarity 0.90)
- f-Domain Adversarial Learning: Theory and Algorithms (ICML 2021 · similarity 0.88)
- Over-parameterized Adversarial Training: An Analysis Overcoming the Curse of Dimensionality (NeurIPS 2020 · similarity 0.88)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.87)
- Rethinking Invariance Regularization in Adversarial Training to Improve Robustness-Accuracy Trade-off (ICLR 2025 · similarity 0.87)
- Stability and Generalization of Adversarial Training for Shallow Neural Networks with Smooth Activation (NeurIPS 2024 · similarity 0.87)
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.