Research map / Federated learning and privacy / Personalized and heterogeneous FL
Seizing Critical Learning Periods in Federated Learning
Read the paperCheck 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 |
|---|---|---|---|
| Recover the full local datasets later after initially training with a partial dataset (ablation) | FedAvg · CIFAR-10, IID and Non-IID | final test accuracy | Hurt |
| Use a smaller share of local data during early training (ablation) | FedAvg · CIFAR-10, IID and Non-IID | final test accuracy | Hurt |
| Use a constant learning rate instead of annealing (ablation) | FedAvg · CIFAR-10, IID and Non-IID | existence of critical learning periods: 0.001 and 0.003 | No clear effect |
| Vary batch size (ablation) | FedAvg · CIFAR-10, IID and Non-IID | existence of critical learning periods | No clear effect |
| Vary weight decay (ablation) | FedAvg · CIFAR-10, IID and Non-IID | shape of critical learning periods | No clear effect |
| Train with all clients during critical periods, then only 60% of clients | FedAvg · IID and Non-IID CIFAR-10 | final test accuracy: 60% | No clear effect |
| Train with all clients during critical periods, then only 60% of clients | FedAvg · IID and Non-IID CIFAR-10 | training time: 60% | Helped |
| Use all local data during critical periods, then only 25% of local data | FedAvg · IID and Non-IID CIFAR-10 | final test accuracy: 25% | No clear effect |
| Use all local data during critical periods, then only 25% of local data | FedAvg · IID and Non-IID CIFAR-10 | training time: 25% | Helped |
| Delay recovery from partial data to the full dataset (ablation) | FedAvg · CIFAR-10, IID and Non-IID | communication rounds required | Hurt |
Closest papers by meaning
- FedImpro: Measuring and Improving Client Update in Federated Learning (ICLR 2024 · similarity 0.87)
- Unlocking the Potential of Model Calibration in Federated Learning (ICLR 2025 · similarity 0.87)
- Federated Learning with Sample-level Client Drift Mitigation (AAAI 2025 · similarity 0.87)
- FLHetBench: Benchmarking Device and State Heterogeneity in Federated Learning (CVPR 2024 · similarity 0.87)
- FedGPS: Statistical Rectification Against Data Heterogeneity in Federated Learning (NeurIPS 2025 · similarity 0.86)
- Adaptive Federated Learning with Auto-Tuned Clients (ICLR 2024 · similarity 0.86)
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.