Research map / Federated learning and privacy / Communication-efficient FL
Single-Round Scalable Analytic Federated Learning
Alan T. L. Bacellar, Mustafa Munir, Felipe M. G. França, Priscila M. V. Lima, Radu Marculescu, Lizy K. John
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 SAFLe non-linear head instead of linear AFL head | ResNet-18 frozen backbone · CIFAR-10 | top-1 accuracy: 80.75 -> 90.73 | Improved |
| use SAFLe instead of multi-round DeepAFL | ResNet-18 · CIFAR-100 | top-1 accuracy: 66.98 -> 70.61 | Improved |
| use SAFLe instead of DeepAFL | ResNet-18 · Tiny-ImageNet | top-1 accuracy: 62.35 -> 64.58 | Improved |
| make data more non-IID (alpha 0.1 -> 0.01) | FedAvg · CIFAR-100 | top-1 accuracy: 56.62 -> 32.99 | Hurt |
| make data more non-IID | SAFLe · CIFAR-100 | top-1 accuracy: 70.61 in all four settings | No clear effect |
| use Binary Overlapping bucketing instead of Integer (ablation) | SAFLe · CIFAR-10 (Bn=8) | top-1 accuracy: 78.15 -> 90.73 | Improved |
| use Binary Overlapping instead of OneHot bucketing (ablation) | SAFLe · CIFAR-100 (Bn=8) | top-1 accuracy: 61.66 -> 70.61 | Improved |
| increase bucket count from 8 to 20 with Integer bucketing (ablation) | SAFLe · CIFAR-10 | top-1 accuracy: 78.15 -> 59.75 | Hurt |
| use many small embeddings (high E, low V) instead of few large ones (ablation) | SAFLe · fixed model size | accuracy | Improved |
| use many small embeddings (ablation) | SAFLe · fixed model size | communication cost | Hurt |
| swap ResNet-18 for ViT-B-16 backbone (ablation) | SAFLe vs AFL · Tiny-ImageNet | top-1 accuracy gain over AFL: 54.67 -> 64.58 (ResNet-18) vs 82.02 -> 82.71 (ViT-B-16) | Mixed |
| use SAFLe instead of DeepAFL | communication · CIFAR-100 at ~67% accuracy | total communication: 146MB -> 70MB | Improved |
Closest papers by meaning
- DeepAFL: Deep Analytic Federated Learning (ICLR 2026 · similarity 0.91)
- AFL: A Single-Round Analytic Approach for Federated Learning with Pre-trained Models (CVPR 2025 · similarity 0.91)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.89)
- Revisiting Ensembling in One-Shot Federated Learning (NeurIPS 2024 · similarity 0.86)
- TurboSVM-FL: Boosting Federated Learning through SVM Aggregation for Lazy Clients (AAAI 2024 · similarity 0.86)
- FedVSSAM: Mitigating Flatness Incompatibility in Sharpness-Aware Federated Learning (NeurIPS 2026 · 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.
