Research map / Federated learning and privacy / Personalized and heterogeneous FL
Self-Aware Personalized Federated Learning
Huili Chen, Jie Ding, Eric William Tramel, Shuang Wu, Anit Kumar Sahu, Salman Avestimehr, Tao Zhang
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 |
|---|---|---|---|
| Use Self-FL instead of FedAvg | FL · FEMNIST top 10% clients | accuracy: 70.48% -> 76.24% | Improved |
| Use Self-FL instead of FedAvg for worst-off clients | FL · FEMNIST worst 10% clients | accuracy: 0% -> 37.01% | Improved |
| Give baselines 40 local steps like Self-FL's cutoff | DITTO, pFedMe, PerFedAvg · FEMNIST | accuracy: 71.33%, 68.87%, 71.08% vs 76.93% | No clear effect |
| Apply only one Self-FL component: adjust local initialization (ablation) | Self-FL · FEMNIST | accuracy | Improved |
| Use sample-size weighted aggregation (FedAvg-w) instead of equal weights (ablation) | FedAvg · Wake-word device A | relative FA: 0.97 -> 8.39 | Hurt |
| Use sample-size weighted aggregation (DITTO-w) instead of equal weights (ablation) | DITTO · Wake-word device A | relative FA: 0.97 -> 8.38 | Hurt |
| Use Self-FL instead of FedAvg-e for global model | FL · Wake-word device A | relative FA: 0.97 -> 0.92 | Improved |
| Use Self-FL instead of PerFedAvg for global model | FL · Wake-word device A | relative FA: 1.06 -> 0.92 | Improved |
| Use Self-FL instead of DITTO-e for global model | FL · Wake-word device E | relative FA: 0.99 -> 1.01 | Hurt |
| Use Self-FL personalized models instead of PerFedAvg | FL · Wake-word device B | relative FA: 1.11 -> 0.91 | Improved |
| Use Self-FL personalized models instead of PerFedAvg | FL · Wake-word device E | relative FA: 0.93 -> 0.99 | Hurt |
| Use Algorithm 2 (augmented) variant instead of Algorithm 1 of Self-FL (ablation) | Self-FL · MNIST/FEMNIST training accuracy | accuracy | No clear effect |
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.
- DITTO — Ditto: Fair and Robust Federated Learning Through Personalization (ICML 2021) · better on 3 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
- Harnessing Heterogeneous Statistical Strength for Personalized Federated Learning via Hierarchical Bayesian Inference (ICML 2025 · similarity 0.90)
- Understanding the Statistical Accuracy-Communication Trade-off in Personalized Federated Learning with Minimax Guarantees (ICML 2025 · similarity 0.90)
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.90)
- FedAS: Bridging Inconsistency in Personalized Federated Learning (CVPR 2024 · similarity 0.89)
- Personalized Federated Learning with Mixture of Models for Adaptive Prediction and Model Fine-Tuning (NeurIPS 2024 · similarity 0.88)
- Knowledge-Aware Parameter Coaching for Personalized Federated Learning (AAAI 2024 · similarity 0.88)
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.
