Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedXDS: Leveraging Model Attribution Methods to counteract Data Heterogeneity in Federated Learning
Maximilian Andreas Hoefler, Karsten Mueller, Wojciech Samek
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 |
|---|---|---|---|
| share attribution-masked, noised features with the server (FedXDS-LRP) | FedAvg · CIFAR-10 K=100, alpha=0.1 | top-1 accuracy: 60.94 -> 80.29 | Improved |
| add FedXDS to FedDyn | FedDyn · CIFAR-10 K=10 | top-1 accuracy: 70.62 -> 84.20 | Improved |
| use LRP rather than Gradient/SmoothGrad/IG as the attribution method (ablation) | FedXDS · CIFAR-10 K=10, alpha=0.1, s=70 | top-1 accuracy: 75.89 (IG) -> 83.46 | Improved |
| increase sparsification level from 60 to 85 (ablation) | FedXDS-LRP · CIFAR-10 K=10, alpha=0.1 | top-1 accuracy: 84.74 -> 81.88 | Hurt |
| increase privacy budget epsilon from 5 to 20 (less noise) (ablation) | FedXDS-LRP · CIFAR-10 K=10, alpha=0.1 | top-1 accuracy: 80.06 -> 83.46 | Improved |
| set knowledge weight lambda to 0.5 instead of 0.0 or 1.0 (ablation) | FedXDS · CIFAR-10 | top-1 accuracy: 72.90 (lambda=0) and 61.18 (lambda=1.0) -> 83.46 | Improved |
| increase warmup rounds (ablation) | FedXDS · CIFAR-10, CIFAR-100, Tiny-ImageNet | accuracy | No clear effect |
| apply attribution masking before adding noise (ablation) | DP feature sharing · membership inference and feature inversion | attack success | Improved |
| FedXLRP versus FedFed on CIFAR-100 K=10 alpha=0.05 and Tiny-ImageNet K=100 alpha=0.1 | FedXLRP · CIFAR-100 / Tiny-ImageNet | top-1 accuracy: 52.07 vs 52.27; 33.18 vs 34.89 | Mixed |
Closest papers by meaning
- FedLMT: Tackling System Heterogeneity of Federated Learning via Low-Rank Model Training with Theoretical Guarantees (ICML 2024 · similarity 0.88)
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · similarity 0.88)
- Bridging Model Heterogeneity in Federated Learning via Uncertainty-based Asymmetrical Reciprocity Learning (ICML 2024 · similarity 0.88)
- Federated Learning via Input-Output Collaborative Distillation (AAAI 2024 · similarity 0.88)
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.87)
- Fed-QSSL: A Framework for Personalized Federated Learning under Bitwidth and Data Heterogeneity (AAAI 2024 · 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.
