Research map / Federated learning and privacy / Personalized and heterogeneous FL
Unsupervised Model Personalization While Preserving Privacy and Scalability: An Open Problem
Check 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 |
|---|---|---|---|
| Estimate IMM importance weights on user data with MAS instead of server data | RACL vs IMM, AlexNet and VGG11 · MITIS CatPrior and TransPrior | accuracy: 66.97 vs 67.39; 47.04 vs 46.51; 77.32 vs 76.77; 53.59 vs 53.49 | Mixed |
| Use MAS rather than FIM importance weights for IMM merging (ablation) | IMM · Numbers, MLP | accuracy: 84.36 vs 87.68 | Hurt |
| Use MAS rather than FIM importance weights for IMM merging (ablation) | IMM · MITIS CatPrior and TransPrior, AlexNet and VGG11 | accuracy: 67.39 vs 67.42; 76.77 vs 76.29; 46.51 vs 46.68; 53.49 vs 53.14 | No clear effect |
| Recompute BN statistics on unlabeled user data with AdaBN (ablation) | RACL and continual-learning baselines · MITIS CatPrior and TransPrior | accuracy: 58.05 -> 58.30; 30.14 -> 30.19; 59.58 -> 59.71; 32.15 -> 32.04 | No clear effect |
| Train BN scale and shift parameters with labeled user data (AdaBN-S) (ablation) | RACL and continual-learning baselines · MITIS CatPrior and TransPrior | accuracy: 58.05 -> 60.68; 30.14 -> 32.82; 59.58 -> 62.43; 32.15 -> 34.80 | Helped |
Closest papers by meaning
- Foundation Cures Personalization: Improving Personalized Models’ Prompt Consistency via Hidden Foundation Knowledge (NeurIPS 2025 · similarity 0.83)
- MMPB: It’s Time for Multi-Modal Personalization (NeurIPS 2025 · similarity 0.83)
- Private Model Personalization Revisited (ICML 2025 · similarity 0.83)
- Ego: Embedding-Guided Personalization of Vision-Language Models (CVPR 2026 · similarity 0.82)
- Domain Generalizable Person Search Using Unreal Dataset (AAAI 2024 · similarity 0.82)
- Perturb a Model, Not an Image: Towards Robust Privacy Protection via Anti-Personalized Diffusion Models (NeurIPS 2025 · similarity 0.82)
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.