Research map / Federated learning and privacy / Personalized and heterogeneous FL
FedRE: A Representation Entanglement Framework for Model-Heterogeneous Federated Learning
Yuan Yao, Lixu Wang, Jiaqi Wu, Jin Song, Simin Chen, Zehua Wang, Zijian Tian, Wei Chen, Huixia Li, Xiaoxiao Li
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 |
|---|---|---|---|
| upload entangled representations to train the global classifier instead of prototypes (FedGH) | model-heterogeneous FL · Average over 6 settings | accuracy: 53.35 -> 57.79 | Improved |
| compute entanglement over prototypes with random weights (RAP) instead of equal weights (VAP) (ablation) | FedRE · CIFAR-10 (PRA) | accuracy: 81.42 -> 82.60 | Improved |
| entangle prototypes rather than raw representations (RAP vs RAR) (ablation) | FedRE · CIFAR-100 (PRA) | accuracy: 43.19 -> 46.36 | Improved |
| entangle all (prototype) items instead of randomly selecting one (ablation) | FedRE · CIFAR-10 (PRA) | accuracy: 79.10 (RSR) -> 82.60 (RAP) | Improved |
| use average pooling for representation mapping instead of max pooling or FC (ablation) | FedRE · CIFAR-100 (PRA) | accuracy: 44.53 (FC), 45.97 (MP) -> 46.36 (AP) | Improved |
| change weight distribution from Uniform to Gaussian or Laplace (ablation) | FedRE RAP · CIFAR-100 (PRA) | accuracy: 46.36 -> 45.51 (Gaussian), 45.84 (Laplace) | No clear effect |
| resample random weights each round instead of fixing them (ablation) | FedRE · Synthetic 2D dataset | accuracy: 41.50 -> 62.00 | Improved |
| resample random weights each round instead of fixing them (ablation) | FedRE · CIFAR-100 (PRA) | accuracy: 45.84 -> 46.36 | Improved |
| upload one entangled representation per client instead of all representations (FedAllRep) | FedRE · TinyImageNet (PRA) | accuracy: 31.20 -> 30.48 | No clear effect |
| add RE during local training (ablation) | FedRE · CIFAR-10 local training time | seconds per round: 5.69 -> 5.78 | No clear effect |
| upload entangled representation instead of prototypes under 10/100 participation, long-tail IF=100 | FedRE vs FedGH · CIFAR-10 PRA 100 clients, IF=100 alpha=0.07 | accuracy: 67.30 -> 67.12 | No clear effect |
| entangle representations before upload instead of uploading prototypes or raw representations | representation inversion attack · TinyImageNet inversion | PSNR: 12.89 -> 9.66 | Improved |
Closest papers by meaning
- Rethinking the Representation in Federated Unsupervised Learning with Non-IID Data (CVPR 2024 · similarity 0.88)
- Federated Model Heterogeneous Matryoshka Representation Learning (NeurIPS 2024 · similarity 0.87)
- FedLMT: Tackling System Heterogeneity of Federated Learning via Low-Rank Model Training with Theoretical Guarantees (ICML 2024 · similarity 0.86)
- Towards Effective Federated Graph Foundation Model via Mitigating Knowledge Entanglement (NeurIPS 2025 · similarity 0.86)
- HPE: Hallucinated Positive Entanglement for Backdoor Attacks in Federated Self-Supervised Learning (NeurIPS 2026 · similarity 0.86)
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · 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.
