Research map / Label-efficient and robust learning / Domain adaptation and generalization
Scaling Unsupervised Multi-Source Federated Domain Adaptation through Group-Wise Discrepancy Minimization
Larissa Reichart, Cem Ata Baykara, Ali Burak Ünal, Harlin Lee, Mete Akgün
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 FACT's single-pair discrepancy with GALA's group discrepancy and weighting | FACT · Digit-18 (9 targets) | average accuracy (%): 87.6 -> 92.9 | Improved |
| use GALA instead of FACT on svhn-xs | FACT · Digit-18 svhn-xs | accuracy (%): 79.1 -> 88.6 | Improved |
| use GALA instead of FACT on synm | FACT · Digit-18 synm | accuracy (%): 79.4 -> 85.8 | Improved |
| use GALA instead of FACT on Digit-Five | FACT · Digit-Five | average accuracy (%): 95.0 -> 95.4 | Improved |
| use GALA instead of KD3A on Office-Caltech10 | KD3A · Office-Caltech10 | average accuracy (%): 97.9 vs 97.8 | No clear effect |
| increase number of sources beyond 9 | FACT · Digit-Five targets | accuracy | Hurt |
| use temperature tau=0.2 instead of 0.8-1.0 (ablation) | GALA · Digit-18 | accuracy | Hurt |
| use tau=3 instead of 0.8-1.0 (ablation) | GALA · Digit-18 | accuracy | Hurt |
| add MDMGB+ weighting to IGD (ablation) | GALA · Digit-18 svhn-xs | convergence/accuracy | Improved |
| add MDMGB weighting to IGD instead of uniform weights (ablation) | IGD · Digit-18 svhn-xs | accuracy | No clear effect |
| add IGD target training to weighting on mnistm (ablation) | GALA · Digit-18 mnistm | accuracy | Improved |
| use GALA instead of KD3A for runtime at 9 sources | KD3A · Digit-Five scalability | seconds per round: 5600.48 -> 22.37 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- KD3A — KD3A: Unsupervised Multi-Source Decentralized Domain Adaptation via Knowledge Distillation (ICML 2021) · better on 2 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Revisiting Source-Free Domain Adaptation: Insights into Representativeness, Generalization, and Variety (CVPR 2025 · similarity 0.89)
- Enhancing Federated Domain Adaptation with Multi-Domain Prototype-Based Federated Fine-Tuning (ICLR 2025 · similarity 0.88)
- Principled Federated Domain Adaptation: Gradient Projection and Auto-Weighting (ICLR 2024 · similarity 0.87)
- UFDA: Universal Federated Domain Adaptation with Practical Assumptions (AAAI 2024 · similarity 0.87)
- Multi-Source Collaborative Gradient Discrepancy Minimization for Federated Domain Generalization (AAAI 2024 · similarity 0.87)
- FedWon: Triumphing Multi-domain Federated Learning Without Normalization (ICLR 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.
