Research map / Federated learning and privacy / Personalized and heterogeneous FL
One-Shot Federated Graph Learning via High-Fidelity Proxies and Transferability-Guided Collaboration
Jia Jiang, Zihan Tan, Wenke Huang, Bin Yang, Mang Ye
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes a one-shot federated graph learning method that builds high-fidelity proxies and uses transferability-guided collaboration to overcome the single-transmission bottleneck.
Problem
One-shot federated graph learning (OFGL) compresses collaboration into a single transmission round under strict bandwidth limits, which restricts the quality and utility of the exchanged knowledge (a "single-transmission bottleneck").
Approach
Generates high-fidelity proxies for the single transmitted knowledge and uses transferability estimates to guide how clients collaborate; the abstract only partially describes this, so specifics are not stated.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use FedHFT instead of strongest one-shot baseline OASIS | one-shot federated graph learning · Cora, CiteSeer, Actor, Amazon-Photo, Roman-Empire, Ogbn-Arxiv | node accuracy (%): Cora 40.05 -> 52.24; Actor 25.51 -> 41.51; Ogbn-Arxiv 28.87 -> 44.50 | Improved |
| use FedHFT instead of OASIS | one-shot federated graph learning · PubMed | node accuracy (%): 65.96 -> 56.98 | Hurt |
| add Fidelity-Driven Proxy Refinement (FPR) (ablation) | FedHFT · Cora, CiteSeer, Actor, Roman-Empire | node accuracy (%): Cora 29.01 -> 41.35 | Improved |
| add Transferability-Steered Personalized Collaboration (TPC) (ablation) | FedHFT · Cora, CiteSeer, Actor, Roman-Empire | node accuracy (%): Cora 29.01 -> 45.25; CiteSeer 27.26 -> 54.19 | Improved |
| combine FPR and TPC (ablation) | FedHFT · Cora | node accuracy (%): 45.25 -> 52.24 | Improved |
| add Fisher gradient matching to FPR (ablation) | FPR · Cora | node accuracy (%): 29.01 -> 37.01 | Improved |
| add class-semantic alignment to FPR (ablation) | FPR · Cora | node accuracy (%): 29.01 -> 39.95 | Improved |
| add class-semantic alignment on top of Fisher matching (ablation) | FPR · Roman-Empire | node accuracy (%): 29.27 -> 30.11 | Improved |
| select top-M channels with M at extremes (0 or 1) rather than 0.5-0.75 (ablation) | FedHFT · Top-M ratio sensitivity | average accuracy | Hurt |
| increase number of clients from 5 to 20 | FedHFT · Cora | node accuracy (%): 62.17 -> 39.81 | Hurt |
Closest papers by meaning
- GHOST: Generalizable One-Shot Federated Graph Learning with Proxy-Based Topology Knowledge Retention (ICML 2025 · similarity 0.91)
- Towards Effective Federated Graph Foundation Model via Mitigating Knowledge Entanglement (NeurIPS 2025 · similarity 0.91)
- FedSSP: Federated Graph Learning with Spectral Knowledge and Personalized Preference (NeurIPS 2024 · similarity 0.89)
- Federated Graph Learning under Domain Shift with Generalizable Prototypes (AAAI 2024 · similarity 0.89)
- Less is More: Federated Graph Learning with Alleviating Topology Heterogeneity from A Causal Perspective (ICML 2025 · similarity 0.89)
- $S^2$FGL: Spatial Spectral Federated Graph Learning (ICML 2025 · similarity 0.88)
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.
