Research map / Federated learning and privacy / Communication-efficient FL
Throughput-Optimal Topology Design for Cross-Silo Federated Learning
Othmane MARFOQ, CHUAN XU, Giovanni Neglia, Richard Vidal
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 the server-client STAR with a designed RING overlay | Cross-silo FL with ResNet-18 · iNaturalist, five evaluated networks | cycle time: 391 -> 118 ms (Gaia); 288 -> 81 ms (AWS North America); 634 -> 109 ms (Géant); 912 -> 103 ms (Exodus); 902 -> 95 ms (Ebo | Improved |
| Replace MATCHA+ with a designed RING overlay | Cross-silo FL with ResNet-18 · iNaturalist, Gaia, AWS North America, Exodus, Ebone | cycle time: 228 -> 118 ms; 124 -> 81 ms; 593 -> 103 ms; 580 -> 95 ms | Improved |
| Replace MATCHA+ with a designed overlay | Cross-silo FL with ResNet-18 · iNaturalist, Géant | cycle time: 452 -> 101 ms (MST) | Improved |
| Use RING rather than STAR with 100 Mbps access links | Cross-silo FL · Exodus network | training time speed-up vs STAR: 72x | Improved |
| Use RING rather than MATCHA+ with 100 Mbps access links | Cross-silo FL · Exodus network | training time speed-up vs MATCHA+: 5.6x | Improved |
| Use RING rather than MST | Cross-silo FL with ResNet-18 · iNaturalist, Géant | cycle time: 101 -> 109 ms | Hurt |
| Increase the number of local computation steps before communication | Cross-silo FL overlays · Experiments varying local steps | throughput | No clear effect |
Closest papers by meaning
- FedCompass: Efficient Cross-Silo Federated Learning on Heterogeneous Client Devices Using a Computing Power-Aware Scheduler (ICLR 2024 · similarity 0.88)
- Efficient Adaptive Federated Optimization (NeurIPS 2025 · similarity 0.86)
- Beyond the Federation: Topology-aware Federated Learning for Generalization to Unseen Clients (ICML 2024 · similarity 0.86)
- Federated Learning under Periodic Client Participation and Heterogeneous Data: A New Communication-Efficient Algorithm and Analysis (NeurIPS 2024 · similarity 0.85)
- FedECADO: A Dynamical System Model of Federated Learning (ICML 2025 · similarity 0.84)
- Exact and Linear Convergence for Federated Learning under Arbitrary Client Participation is Attainable (NeurIPS 2025 · similarity 0.84)
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.
