Summary of Federated Graph Learning with Structure Proxy Alignment, by Xingbo Fu et al.
Federated Graph Learning with Structure Proxy Alignment
by Xingbo Fu, Zihan Chen, Binchi Zhang, Chen Chen, Jundong Li
First submitted to arxiv on: 18 Aug 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); 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) aims to learn graph learning models over graph data distributed in multiple data owners, which has been applied in various applications such as social recommendation and financial fraud detection. The paper highlights the challenges of FGL due to data heterogeneity, where label distribution may vary significantly across clients. Specifically, local objectives diverge for node-level tasks, especially node classification. Additionally, nodes from minority classes are more likely to have biased neighboring information, impeding expressive node embeddings with Graph Neural Networks (GNNs). To address this challenge, the authors propose FedSpray, a novel FGL framework that learns local class-wise structure proxies and aligns them for global structure proxies. This enables the generation of unbiased soft targets and regularizes local training of GNN models. The authors conduct experiments on four datasets and demonstrate the superiority of FedSpray compared to other baselines. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated Graph Learning (FGL) is a way for machines to learn from different pieces of graph data without sharing those data. This is useful in applications like social recommendation and financial fraud detection. One problem with FGL is that the data is not always the same, which makes it hard for the model to learn. For example, some clients might have mostly one type of node, while others have a mix. This makes it hard for the model to work well. Another challenge is when there are only a few nodes from a certain class on a client, but those nodes are surrounded by biased information that makes them look more like other classes. To solve these problems, the authors created a new FGL framework called FedSpray. It learns local patterns and aligns them to get a global understanding of the data. This helps create unbiased targets for training GNN models. The authors tested their approach on several datasets and found that it worked better than other methods. |
Keywords
» Artificial intelligence » Classification » Gnn
