Research map / Federated learning and privacy / Communication-efficient FL
Sharp Bounds for Sequential Federated Learning on Heterogeneous Data
Yipeng Li, Xinchen Lyu
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 sequential FL instead of parallel FL | CNN federated training · Fashion-MNIST (C=1) | test accuracy: 85.77 -> 87.60 | Improved |
| use sequential FL instead of parallel FL | CNN federated training · Fashion-MNIST (C=2) | test accuracy: 88.55 -> 89.66 | Improved |
| use sequential FL instead of parallel FL | CNN federated training · Fashion-MNIST (C=5) | test accuracy: 90.70 -> 90.92 | Improved |
| use sequential FL instead of parallel FL | VGG-9 federated training · CIFAR-10 (C=1) | test accuracy: 73.84 -> 81.05 | Improved |
| use sequential FL instead of parallel FL | VGG-9 federated training · CIFAR-10 (C=2) | test accuracy: 78.99 -> 83.34 | Improved |
| use sequential FL instead of parallel FL | VGG-9 federated training · CIFAR-10 (C=5) | test accuracy: 83.47 -> 84.73 | Improved |
| use sequential FL instead of parallel FL | VGG-9 federated training · CINIC-10 (C=1) | test accuracy: 52.27 -> 61.52 | Improved |
| use sequential FL instead of parallel FL | VGG-9 federated training · CINIC-10 (C=2) | test accuracy: 61.96 -> 67.31 | Improved |
| use sequential FL instead of parallel FL | VGG-9 federated training · CINIC-10 (C=5) | test accuracy: 68.45 -> 70.82 | Improved |
| use sequential FL instead of parallel FL on one-dimensional quadratics with learning rate of SFL set to half of PFL | SFL vs PFL · Quadratic functions (10 groups, M=2, K=10) | convergence of objective | Mixed |
| use sequential FL instead of parallel FL under high heterogeneity (C=1) | SFL vs PFL · Logistic regression a9a and w8a (M=1000) | training loss | Improved |
| use sequential FL instead of parallel FL under low heterogeneity (C=2) | SFL vs PFL · Logistic regression a9a and w8a (M=1000) | training loss | No clear effect |
Closest papers by meaning
- Convergence Analysis of Split Federated Learning on Heterogeneous Data (NeurIPS 2024 · similarity 0.92)
- A New Theoretical Perspective on Data Heterogeneity in Federated Optimization (ICML 2024 · similarity 0.90)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.88)
- Decentralized Sporadic Federated Learning: A Unified Algorithmic Framework with Convergence Guarantees (ICLR 2025 · similarity 0.87)
- SemiDFL: A Semi-Supervised Paradigm for Decentralized Federated Learning (AAAI 2025 · similarity 0.87)
- FedGPS: Statistical Rectification Against Data Heterogeneity in Federated Learning (NeurIPS 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.
