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) is a promising approach for applications withnonEuclidean data. However, training GNNs on large scale graphs with hundredsof millions nodes is both resource and time consuming. Different from DNNs,GNNs usually have larger memory footprints, and thus the GPU memory capacityand PCIe bandwidth are the main resource bottlenecks in GNN training. Toaddress this problem, we present BiFeat: a graph feature quantizationmethodology to accelerate GNN training by significantly reducing the memoryfootprint and PCIe bandwidth requirement so that GNNs can take full advantageof GPU computing capabilities. Our key insight is that unlike DNN, GNN is lessprone to the information loss of input features caused by quantization. Weidentify the main accuracy impact factors in graph feature quantization andtheoretically prove that BiFeat training converges to a network where the lossis within $\epsilon$ of the optimal loss of uncompressed network. We performextensive evaluation of BiFeat using several popular GNN models and datasets,including GraphSAGE on MAG240M, the largest public graph dataset. The resultsdemonstrate that BiFeat achieves a compression ratio of more than 30 andimproves GNN training speed by 200%-320% with marginal accuracy loss. Inparticular, BiFeat achieves a record by training GraphSAGE on MAG240M withinone hour using only four GPUs.

BiFeat: Supercharge GNN Training via Graph Feature Quantization