Research map / Federated learning and privacy / Personalized and heterogeneous FL
Handling Data Heterogeneity via Architectural Design for Federated Visual Recognition
Sara Pieri, Jose Renato Restom, Samuel Horváth, Hisham Cholakkal
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 ResNet-50 with CAFormer-S18 | FedAvg on CIFAR-10 Split-3 · CIFAR-10 Split-3 | accuracy: 59.9 \pm 1.87 -> 97.1 \pm 0.14 | Improved |
| Replace ResNet-50 with ViT-S | federated visual recognition · GLD-23K | accuracy: 54.3 \pm 0.48 -> 76.6 \pm 0.69 | Improved |
| Replace ResNet-50 with CAFormer-S18 | federated visual recognition · Fed-ISIC2019 | accuracy: 67.6 \pm 1.59 -> 83.1 \pm 0.49 | Improved |
| Replace BatchNorm with LayerNorm (ablation) | CoAtNet-0 · CIFAR-10 Split-3 | accuracy: 20.9 \pm 0.45 -> 94.2 \pm 0.06 | Improved |
| Replace BatchNorm with LayerNorm (ablation) | PoolFormer-S12 · CIFAR-10 Split-3 | accuracy: 51.5 \pm 0.32 -> 91.5 \pm 0.13 | Improved |
| Replace BatchNorm with LayerNorm (ablation) | ResNet-50 · CIFAR-10 Split-3 | accuracy: 59.9\pm 1.87 -> 72.2 \pm 3.06 | Improved |
| Use FedBN while retaining BatchNorm (ablation) | ResNet-50 · CIFAR-10 Split-3 | accuracy: 59.9\pm 1.87 -> 59.6 \pm 1.40 | No clear effect |
| Replace BatchNorm with LayerNorm (ablation) | PoolFormer-S12 · Fed-ISIC2019 | accuracy: 64.2 \pm 0.53 -> 83.2 \pm 0.63 | Improved |
| Use GroupNorm instead of LayerNorm (ablation) | PoolFormer-S12 · CIFAR-10 Split-3 | accuracy: 91.5 \pm 0.13 -> 81.5 \pm 0.29 | Hurt |
| Use GroupNorm instead of LayerNorm (ablation) | ResNet-50 · CIFAR-10 Split-3 | accuracy: 72.2 \pm 3.06 -> 89.3\pm 0.59 | Improved |
| Use FedAVGM instead of FedAVG (ablation) | ResNet-50 · CIFAR-10 Split-3 | accuracy: 59.9 \pm 1.87 -> 80.2 \pm 0.49 | Improved |
| Use SCAFFOLD instead of FedAVG (ablation) | CAFormer-S18 · Fed-ISIC2019 | accuracy: 83.1 \pm 0.49 -> 77.1 \pm 0.61 | Hurt |
Closest papers by meaning
- F^3OCUS - Federated Finetuning of Vision-Language Foundation Models with Optimal Client Layer Updating Strategy via Multi-objective Meta-Heuristics (CVPR 2025 · similarity 0.87)
- CLIP-Guided Federated Learning on Heterogeneity and Long-Tailed Data (AAAI 2024 · similarity 0.87)
- FedQS: Optimizing Gradient and Model Aggregation for Semi-Asynchronous Federated Learning (NeurIPS 2025 · similarity 0.86)
- FedDAT: An Approach for Foundation Model Finetuning in Multi-Modal Heterogeneous Federated Learning (AAAI 2024 · similarity 0.86)
- FORLA: Federated Object-centric Representation Learning with Slot Attention (NeurIPS 2025 · similarity 0.86)
- Global and Local Prompts Cooperation via Optimal Transport for Federated Learning (CVPR 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.
