Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedVeer: Self-Adaptive Skew Estimation for Robust Federated Learning
Yun Xin, Bangqi Pan, Jianfeng Lu, Shuqin Cao, Gang Li, Guanghui Wen
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 |
|---|---|---|---|
| add self-adaptive kernel density module (k-free neighborhood via max-margin) (ablation) | FedVeer · FEMNIST | accuracy: 0.463 -> 0.636 | Improved |
| add self-adaptive kernel density module (ablation) | FedAvg-style aggregation · FEMNIST | accuracy: 0.377 -> 0.608 | Improved |
| add Kalman filter margin stabilization (ablation) | FedVeer · FEMNIST | accuracy: 0.608 -> 0.636 | Improved |
| add Kalman filter margin stabilization (ablation) | FedVeer · CIFAR-10 | accuracy: 0.588 -> 0.591 | Improved |
| add Kalman filter alone without adaptive kernel (ablation) | FedAvg-style aggregation · CIFAR-10 | accuracy: 0.567 -> 0.568 | No clear effect |
| vary neighborhood size k from 5 to 15 (ablation) | FedVeer · CEKA datasets | accuracy: audiology 0.545 (k=10) to 0.742 (k=8) | Mixed |
| shrink kernel bandwidth h from x to x/4 (ablation) | FedVeer · CIFAR-10 | accuracy: 0.511 -> 0.238 (k=5) | Hurt |
| shrink kernel bandwidth h from x to x/4 (ablation) | FedVeer · Fashion-MNIST | accuracy: 0.739 -> 0.735 (k=5) | No clear effect |
| increase Kalman forgetting factor lambda from 0.3 to 0.9 (ablation) | Kalman filter · MNIST | error reduction ratio (x10^-3): 0.071 -> 0.195 | Improved |
| decrease Kalman process noise alpha from 0.020 to 0.001 (ablation) | Kalman filter · MNIST | error reduction ratio (x10^-3): 0.110 -> 0.351 | Improved |
| increase observation noise beta from 0.01 to 0.10 (ablation) | Kalman filter · MNIST | error reduction ratio (x10^-3): 0.110 -> 0.300 | Improved |
| use FedVeer instead of best baseline TPFL on CEKA average | federated aggregation · CEKA datasets (34) | accuracy: 0.840 -> 0.866 | Improved |
Closest papers by meaning
- Fair Federated Learning under Domain Skew with Local Consistency and Domain Diversity (CVPR 2024 · similarity 0.88)
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · similarity 0.88)
- Efficient Federated Learning against Heterogeneous and Non-stationary Client Unavailability (NeurIPS 2024 · similarity 0.87)
- FedAdaVR: Adaptive Variance Reduction for Robust Federated Learning under Limited Client Participation (NeurIPS 2026 · similarity 0.87)
- A Simple Data Augmentation for Feature Distribution Skewed Federated Learning (CVPR 2025 · similarity 0.86)
- FedImpro: Measuring and Improving Client Update in Federated Learning (ICLR 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.
