Summary of Federated Graph Learning with Graphless Clients, by Xingbo Fu et al.
Federated Graph Learning with Graphless Clients
by Xingbo Fu, Song Wang, Yushun Dong, Binchi Zhang, Chen Chen, Jundong Li
First submitted to arxiv on: 13 Nov 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Distributed, Parallel, and Cluster Computing (cs.DC)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Federated Graph Learning (FGL) tackles training machine learning models, such as Graph Neural Networks (GNNs), for multiple clients with unique graph data. Existing methods assume each client has node features and graph structure. However, in real-world scenarios, some clients have only node features, while others are “graphless” clients. This leads to a novel problem: how to jointly train a model over distributed graph data with graphless clients? The paper proposes FedGLS, a framework that tackles this issue by devising local graph learners on each client, transferring structure knowledge via GNN models and feature encoders. During training, feature encoders retain local graph structure knowledge, which is transferred among clients in global updates. Our extensive experiments show the superiority of FedGLS over five baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine you have many friends with their own special networks, like social media or phone contacts. Each friend has some information about their connections, but some friends might not know much about the structure of those connections. The problem is: how do we train a model to work on all these networks, even if some friends don’t have that extra structural information? This paper proposes a new way to solve this problem called FedGLS. It helps local learners understand their own network structures and shares knowledge among friends. We tested it with many different models and showed that it works better than other approaches. |
Keywords
* Artificial intelligence * Gnn * Machine learning
