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) are powerful tools for learning from graph dataand are widely used in various applications such as social networkrecommendation, fraud detection, and graph search. The graphs in theseapplications are typically large, usually containing hundreds of millions ofnodes. Training GNN models on such large graphs efficiently remains a bigchallenge. Despite a number of sampling-based methods have been proposed toenable mini-batch training on large graphs, these methods have not been provedto work on truly industry-scale graphs, which require GPUs or mixed-CPU-GPUtraining. The state-of-the-art sampling-based methods are usually not optimizedfor these real-world hardware setups, in which data movement between CPUs andGPUs is a bottleneck. To address this issue, we propose Global NeighborhoodSampling that aims at training GNNs on giant graphs specifically formixed-CPU-GPU training. The algorithm samples a global cache of nodesperiodically for all mini-batches and stores them in GPUs. This global cacheallows in-GPU importance sampling of mini-batches, which drastically reducesthe number of nodes in a mini-batch, especially in the input layer, to reducedata copy between CPU and GPU and mini-batch computation without compromisingthe training convergence rate or model accuracy. We provide a highly efficientimplementation of this method and show that our implementation outperforms anefficient node-wise neighbor sampling baseline by a factor of 2X-4X on giantgraphs. It outperforms an efficient implementation of LADIES with small layersby a factor of 2X-14X while achieving much higher accuracy than LADIES.We alsotheoretically analyze the proposed algorithm and show that with cached nodedata of a proper size, it enjoys a comparable convergence rate as theunderlying node-wise sampling method.

Global Neighbor Sampling for Mixed CPU-GPU Training on Giant Graphs