Research map / Federated learning and privacy / Communication-efficient FL
On the Convergence of Communication-Efficient Local SGD for 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 |
|---|---|---|---|
| Use error-compensated double compression instead of full-precision communication | Local SGD · CIFAR-10 | Top-1 test accuracy (%): 92.68 -> 92.76 (p=1); 92.65 -> 92.62 (p=4); 92.47 -> 92.58 (p=8); 92.86 -> 92.42 (p=16) | No clear effect |
| Use error-compensated double compression instead of full-precision communication | Momentum Local SGD · ImageNet | Top-1 test accuracy (%): 76.04 -> 75.88 (p=1); 75.81 -> 76.12 (p=4); 76.01 -> 76.06 (p=8); 76.01 -> 75.94 (p=16) | No clear effect |
| Increase the communication period from p=1 to p=16 (ablation) | Alg1 · CIFAR-10 | Top-1 test accuracy (%): 92.76 -> 92.42 | No clear effect |
| Increase the communication period from p=1 to p=16 (ablation) | Alg2 · ImageNet | Top-1 test accuracy (%): 75.88 -> 75.94 | No clear effect |
| Compress both worker-to-server and server-to-worker communication | Alg1 versus Qsparse-local-SGD · CIFAR-10 | communication cost | Helped |
| Use top-10% sparsification with error compensation | Local SGD · CIFAR-10 | Top-1 test accuracy (%): 92.68 -> 92.76 (p=1); 92.65 -> 92.62 (p=4); 92.47 -> 92.58 (p=8); 92.86 -> 92.42 (p=16) | No clear effect |
| Use top-10% sparsification with error compensation | Momentum Local SGD · ImageNet | Top-1 test accuracy (%): 76.04 -> 75.88 (p=1); 75.81 -> 76.12 (p=4); 76.01 -> 76.06 (p=8); 76.01 -> 75.94 (p=16) | No clear effect |
Closest papers by meaning
- Federated Learning under Arbitrary Communication Patterns (ICML 2021 · similarity 0.93)
- Hybrid Local SGD for Federated Learning with Heterogeneous Communications (ICLR 2022 · similarity 0.92)
- LOSCAR-SGD: Local SGD with Communication-Computation Overlap and Delay-Corrected Sparse Model Averaging (NeurIPS 2026 · similarity 0.92)
- Communication-Efficient Federated Learning with Accelerated Client Gradient (CVPR 2024 · similarity 0.88)
- Low Precision Local Training is Enough for Federated Learning (NeurIPS 2024 · similarity 0.87)
- FedSMU: Communication-Efficient and Generalization-Enhanced Federated Learning through Symbolic Model Updates (ICML 2025 · 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.