English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Recently, Federated Graph Learning (FGL) has attracted significant attentionas a distributed framework based on graph neural networks, primarily due to itscapability to break data silos. Existing FGL studies employ community split onthe homophilous global graph by default to simulate federated semi-supervisednode classification settings. Such a strategy assumes the consistency oftopology between the multi-client subgraphs and the global graph, whereconnected nodes are highly likely to possess similar feature distributions andthe same label. However, in real-world implementations, the varyingperspectives of local data engineering result in various subgraph topologies,posing unique heterogeneity challenges in FGL. Unlike the well-known labelNon-independent identical distribution (Non-iid) problems in federatedlearning, FGL heterogeneity essentially reveals the topological divergenceamong multiple clients, namely homophily or heterophily. To simulate and handlethis unique challenge, we introduce the concept of structure Non-iid split andthen present a new paradigm called \underline{Ada}ptive \underline{F}ederated\underline{G}raph \underline{L}earning (AdaFGL), a decoupled two-steppersonalized approach. To begin with, AdaFGL employs standard multi-clientfederated collaborative training to acquire the federated knowledge extractorby aggregating uploaded models in the final round at the server. Then, eachclient conducts personalized training based on the local subgraph and thefederated knowledge extractor. Extensive experiments on the 12 graph benchmarkdatasets validate the superior performance of AdaFGL over state-of-the-artbaselines. Specifically, in terms of test accuracy, our proposed AdaFGLoutperforms baselines by significant margins of 3.24\% and 5.57\% on communitysplit and structure Non-iid split, respectively.

AdaFGL: A New Paradigm for Federated Node Classification with Topology Heterogeneity