Research map / Label-efficient and robust learning / Domain adaptation and generalization
Deep Adversarial Attention Alignment for Unsupervised Domain Adaptation: the Benefit of Target Expectation Maximization
Guoliang Kang, Liang Zheng, Yan Yan, Yi Yang
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 |
|---|---|---|---|
| add attention alignment loss L_AT (ablation) | EM target training · MNIST->MNIST-M | accuracy (%): 93.5 -> 95.6 | Improved |
| add attention alignment loss L_AT to a CNN trained on S+Tf (ablation) | CNN · MNIST->MNIST-M | accuracy (%): 78.5 -> 85.7 | Improved |
| add synthetic target data to source network training (ablation) | CNN · MNIST->MNIST-M | accuracy (%): 45.6 -> 75.0 | Improved |
| mix source and synthetic target data (ablation) | CNN · MNIST->MNIST-M | accuracy (%): 75.0 -> 78.5 | Improved |
| add attention alignment loss L_AT (ablation) | EM, Office-31 4 tasks · D->A | accuracy (%): 72.3 -> 74.3 | Improved |
| add attention alignment loss L_AT (ablation) | EM, Office-31 4 tasks · W->A | accuracy (%): 72.2 -> 73.9 | Improved |
| add attention alignment loss L_AT (ablation) | EM, Office-31 average of 6 tasks · Office-31 | accuracy (%): 86.3 -> 87.2 | Improved |
| add asynchronous update of M_post (full vs EM-A) (ablation) | EM · Office-31 4 tasks avg | accuracy (%): 64.4 -> 79.7 | Improved |
| add learning-rate schedule re-initialization (full vs EM-C) (ablation) | EM · Office-31 4 tasks avg | accuracy (%): 78.6 -> 79.7 | Improved |
| add filtering of noisy data (full vs EM-B) (ablation) | EM · Office-31 4 tasks avg | accuracy (%) | Improved |
| use L2 attention discrepancy instead of L1 (ablation) | L_AT · Office-31 4 tasks avg | accuracy (%) | Improved |
| use L2 attention discrepancy instead of JMMD (ablation) | L_AT · Office-31 4 tasks avg | accuracy (%): 78.1 -> 80.9 | Improved |
Closest papers by meaning
- Co-regularized Alignment for Unsupervised Domain Adaptation (NeurIPS 2018 · similarity 0.92)
- MultiDIAL: Domain Alignment Layers for (Multisource) Unsupervised Domain Adaptation (TPAMI 2020 · similarity 0.92)
- Transferable Attention for Domain Adaptation (AAAI 2019 · similarity 0.90)
- Enhancing Domain Adaptation through Prompt Gradient Alignment (NeurIPS 2024 · similarity 0.90)
- Prompt-Based Distribution Alignment for Unsupervised Domain Adaptation (AAAI 2024 · similarity 0.90)
- Toward Improving Robustness and Accuracy in Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.90)
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.
