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

Graph convolutional networks (GCNs) have been widely adopted for graphrepresentation learning and achieved impressive performance. For larger graphsstored separately on different clients, distributed GCN training algorithmswere proposed to improve efficiency and scalability. However, existing methodsdirectly exchange node features between different clients, which results indata privacy leakage. Federated learning was incorporated in graph learning totackle data privacy, while they suffer from severe performance drop due tonon-iid data distribution. Besides, these approaches generally involve heavycommunication and memory overhead during the training process. In light ofthese problems, we propose a Privacy-Preserving Subgraph sampling baseddistributed GCN training method (PPSGCN), which preserves data privacy andsignificantly cuts back on communication and memory overhead. Specifically,PPSGCN employs a star-topology client-server system. We firstly sample a localnode subset in each client to form a global subgraph, which greatly reducescommunication and memory costs. We then conduct local computation on eachclient with features or gradients of the sampled nodes. Finally, all clientssecurely communicate with the central server with homomorphic encryption tocombine local results while preserving data privacy. Compared with federatedgraph learning methods, our PPSGCN model is trained on a global graph to avoidthe negative impact of local data distribution. We prove that our PPSGCNalgorithm would converge to a local optimum with probability 1. Experimentresults on three prevalent benchmarks demonstrate that our algorithmsignificantly reduces communication and memory overhead while maintainingdesirable performance. Further studies not only demonstrate the fastconvergence of PPSGCN, but discuss the trade-off between communication andlocal computation cost as well.

PPSGCN: A Privacy-Preserving Subgraph Sampling Based Distributed GCN Training Method