Research map / Label-efficient and robust learning / Domain adaptation and generalization
Inferring Latent Domains for Unsupervised Deep Domain Adaptation
Massimiliano Mancini, Lorenzo Porzi, Samuel Rota Bulò, Barbara Caputo B, Elisa Ricci
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 domain balance entropy loss (lambda_B>0) (ablation) | latent-domain mDA network · PACS multi-source multi-target (mean) | accuracy: 82.6 -> 83.3 | Improved |
| add domain balance entropy loss (lambda_B>0) (ablation) | latent-domain mDA network · PACS single target (mean) | accuracy: 85.3 -> 85.4 | Improved |
| add domain balance entropy loss (lambda_B>0) (ablation) | latent-domain mDA network · Digits M-Mm to U | accuracy: 82.5 -> 82.4 | No clear effect |
| add domain balance entropy loss (lambda_B>0) (ablation) | latent-domain mDA network · Digits-five (mean) | accuracy: 75.8 -> 76.4 | Improved |
| add domain balance entropy loss (lambda_B>0) (ablation) | latent-domain mDA network · Office-31 multi-source (mean) | accuracy: 83.6 -> 83.2 | No clear effect |
| add domain balance entropy loss (lambda_B>0) (ablation) | latent-domain mDA network · Office-31 latent domain discovery comparison (mean) | accuracy: 83.9 -> 84.8 | Improved |
| add domain balance entropy loss (lambda_B>0) (ablation) | latent-domain mDA network · Office-Caltech latent domain discovery comparison (mean) | accuracy: 91.8 -> 92.5 | Improved |
| learn domain assignments instead of random assignment (ablation) | latent-domain mDA network · PACS multi-source multi-target (mean) | accuracy: 79.1 -> 83.3 | Improved |
| add losses on domain prediction branch (lambda_E, lambda_B) (ablation) | latent-domain mDA network · PACS multi-source multi-target (mean) | accuracy: 80.2 -> 83.3 | Improved |
| add entropy loss on unlabeled target classification (lambda_C) (ablation) | latent-domain mDA network · PACS multi-source multi-target (mean) | accuracy: 76.4 -> 82.6 | Improved |
| vary number of discovered source domains k from 2 to 5 (ablation) | latent-domain mDA network · Digits M-Mm to U | accuracy: 82.5, 82.2, 82.7, 82.4 | No clear effect |
| replace DIAL with latent domain discovery mDA layers | DIAL · PACS single target (mean) | accuracy: 84.2 -> 85.4 | Improved |
Closest papers by meaning
- Boosting Domain Adaptation by Discovering Latent Domains (CVPR 2018 · similarity 0.93)
- Can We Evaluate Domain Adaptation Models Without Target-Domain Labels? (ICLR 2024 · similarity 0.92)
- Moment Matching for Multi-Source Domain Adaptation (ICCV 2019 · similarity 0.91)
- Towards Uncovering the Intrinsic Data Structures for Unsupervised Domain Adaptation Using Structurally Regularized Deep Clustering (TPAMI 2021 · similarity 0.91)
- Incorporating Pre-Training Data Matters in Unsupervised Domain Adaptation (TPAMI 2025 · similarity 0.91)
- Make the U in UDA Matter: Invariant Consistency Learning for Unsupervised Domain Adaptation (NeurIPS 2023 · similarity 0.91)
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.
