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

A key performance bottleneck when training graph neural network (GNN) modelson large, real-world graphs is loading node features onto a GPU. Due to limitedGPU memory, expensive data movement is necessary to facilitate the storage ofthese features on alternative devices with slower access (e.g. CPU memory).Moreover, the irregularity of graph structures contributes to poor datalocality which further exacerbates the problem. Consequently, existingframeworks capable of efficiently training large GNN models usually incur asignificant accuracy degradation because of the inevitable shortcuts involved.To address these limitations, we instead propose ReFresh, a general-purpose GNNmini-batch training framework that leverages a historical cache for storing andreusing GNN node embeddings instead of re-computing them through fetching rawfeatures at every iteration. Critical to its success, the corresponding cachepolicy is designed, using a combination of gradient-based and stalenesscriteria, to selectively screen those embeddings which are relatively stableand can be cached, from those that need to be re-computed to reduce estimationerrors and subsequent downstream accuracy loss. When paired with complementarysystem enhancements to support this selective historical cache, ReFresh is ableto accelerate the training speed on large graph datasets such asogbn-papers100M and MAG240M by 4.6x up to 23.6x and reduce the memory access by64.5% (85.7% higher than a raw feature cache), with less than 1% influence ontest accuracy.

ReFresh: Reducing Memory Access from Exploiting Stable Historical Embeddings for Graph Neural Network Training