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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Graph neural networks (GNNs) have extended the success of deep neuralnetworks (DNNs) to non-Euclidean graph data, achieving ground-breakingperformance on various tasks such as node classification and graph propertyprediction. Nonetheless, existing systems are inefficient to train large graphswith billions of nodes and edges with GPUs. The main bottlenecks are theprocess of preparing data for GPUs - subgraph sampling and feature retrieving.This paper proposes BGL, a distributed GNN training system designed to addressthe bottlenecks with a few key ideas. First, we propose a dynamic cache engineto minimize feature retrieving traffic. By a co-design of caching policy andthe order of sampling, we find a sweet spot of low overhead and high cache hitratio. Second, we improve the graph partition algorithm to reducecross-partition communication during subgraph sampling. Finally, carefulresource isolation reduces contention between different data preprocessingstages. Extensive experiments on various GNN models and large graph datasetsshow that BGL significantly outperforms existing GNN training systems by 20.68xon average.

BGL: GPU-Efficient GNN Training by Optimizing Graph Data I/O and Preprocessing