Research map / Federated learning and privacy / Personalized and heterogeneous FL
Federated Distillation for Whole Slide Image via Gaussian-Mixture Feature Alignment and Curriculum Integration
Luru Jing, Cong Cong, Yanyuan Chen, Yongzhi Cao
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 feature-level distillation (FDD) instead of pixel-level (ablation) | FedHD · CAMELYON16 / CAMELYON17 / TCGA-IDH | accuracy (%): 88.7 -> 89.7 / 77.2 -> 78.8 / 80.5 -> 82.1 | Improved |
| add Gaussian-mixture alignment (GMA) alone (ablation) | FedHD · CAMELYON16 | accuracy (%): 88.7 -> 89.4 | Improved |
| add Gaussian-mixture alignment (GMA) alone (ablation) | FedHD · TCGA-IDH | accuracy (%): 80.5 -> 83.4 | Improved |
| add one-to-one distillation (O2O) to FDD (ablation) | FedHD · CAMELYON17 | accuracy (%): 78.8 -> 81.2 | Improved |
| add GMA on top of FDD&O2O (ablation) | FedHD · CAMELYON17 | accuracy (%): 81.2 -> 81.9 | Improved |
| add curriculum-based federation (CBF) alone (ablation) | FedHD · CAMELYON17 | accuracy (%): 77.2 -> 80.7 | Improved |
| add CBF on top of FDD&O2O&GMA (ablation) | FedHD · CAMELYON16 | accuracy (%): 91.2 -> 91.2 | No clear effect |
| add CBF on top of FDD&O2O&GMA (ablation) | FedHD · TCGA-IDH | accuracy (%): 84.6 -> 84.8 | Improved |
| FedHD vs strongest baseline FedWSIDD on heterogeneous models | FedHD · CAMELYON16 | accuracy (%): 88.7 -> 91.2 | Improved |
| FedHD vs FedWSIDD privacy under membership inference | FedHD · TCGA-IDH | LiRA mean AUC (%): 56.1 -> 52.3 | Improved |
| feature-level distillation instead of image distillation | FedHD · TCGA-IDH client C1 | distillation time: 10-12 hours -> approximately 1 hour | Improved |
| enable O2O distillation | FedHD · TCGA-IDH client C1 | communication payload: approximately 39 MB -> approximately 1.19 GB | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- HistoFS — HistoFS: Non-IID Histopathologic Whole Slide Image Classification via Federated Style Transfer with RoI-Preserving (CVPR 2025) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- HistoFS: Non-IID Histopathologic Whole Slide Image Classification via Federated Style Transfer with RoI-Preserving (CVPR 2025 · similarity 0.86)
- PathVQ: Reforming Computational Pathology Foundation Model for Whole Slide Image Analysis via Vector Quantization (NeurIPS 2025 · similarity 0.86)
- FedVCK: Non-IID Robust and Communication-Efficient Federated Learning via Valuable Condensed Knowledge for Medical Image Analysis (AAAI 2025 · similarity 0.85)
- MH-pFLID: Model Heterogeneous personalized Federated Learning via Injection and Distillation for Medical Data Analysis (ICML 2024 · similarity 0.85)
- Data-Free Black-Box Federated Learning via Zeroth-Order Gradient Estimation (AAAI 2025 · similarity 0.85)
- FedPIA – Permuting and Integrating Adapters Leveraging Wasserstein Barycenters for Finetuning Foundation Models in Multi-Modal Federated Learning (AAAI 2025 · similarity 0.85)
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.
