Research map / Federated learning and privacy / Communication-efficient FL
FedGCN: Convergence-Communication Tradeoffs in Federated Training of Graph Convolutional Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 1-hop cross-client neighbor information communication (pre-training only) (ablation) | FedGCN · Cora, 10 clients, beta=1 | test accuracy: 0.6502 -> 0.81 | Helped |
| add 1-hop communication (ablation) | FedGCN · Ogbn-Arxiv, 10 clients, beta=1 | test accuracy: 0.5981 -> 0.7101 | Helped |
| add 1-hop communication (ablation) | FedGCN · Ogbn-Products, 5 clients, beta=100 | test accuracy: 0.658 -> 0.7057 | Helped |
| increase from 1-hop to 2-hop communication, cross-silo (10 clients) (ablation) | FedGCN · Cora, Citeseer, Arxiv, Products | test accuracy: Cora beta=10000: 0.8009 -> 0.8087 | No clear effect |
| increase from 1-hop to 2-hop with many clients (cross-device) (ablation) | FedGCN · Cora, varying number of clients | test accuracy | Helped |
| use FedGCN (1-hop) instead of FedSage+ | federated GCN · Cora, beta=1 | test accuracy: 0.8026 -> 0.81 | Helped |
| use FedGCN (1-hop) instead of BDS-GCN | federated GCN · Ogbn-Arxiv, beta=1 | test accuracy: 0.6769 -> 0.7101 | Helped |
| one-time pre-training communication instead of per-round communication | FedGCN vs BDS-GCN and FedSage+ · Ogbn-ArXiv | communication cost: at least 100x less | Helped |
| non-i.i.d. partition instead of i.i.d. for 0-hop | FedGCN (0-hop) · Cora, 10 clients | test accuracy: 0.5992 -> 0.6502 | Helped |
| use Boolean Packing for HE (ablation) | CKKS secure aggregation · 1M array | communication size: 70 MB -> 8 MB | Helped |
Closest papers by meaning
- Virtual Nodes Can Help: Tackling Distribution Shifts in Federated Graph Learning (AAAI 2025 · similarity 0.90)
- Federated Graph Condensation with Information Bottleneck Principles (AAAI 2025 · similarity 0.90)
- Federated Graph-Level Clustering Network (AAAI 2025 · similarity 0.89)
- FedSPA: Generalizable Federated Graph Learning under Homophily Heterogeneity (CVPR 2025 · similarity 0.89)
- Less is More: Federated Graph Learning with Alleviating Topology Heterogeneity from A Causal Perspective (ICML 2025 · similarity 0.89)
- GRAIN: Exact Graph Reconstruction from Gradients (ICLR 2025 · similarity 0.89)
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.