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/

Zendy Plus

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

Zendy Tools

Zendy Open

Dynamic Graph Neural Networks (DGNNs) have been broadly applied in variousreal-life applications, such as link prediction and pandemic forecast, tocapture both static structural information and temporal characteristics fromdynamic graphs. Combining both time-dependent and -independent components,DGNNs manifest substantial parallel computation and data reuse potentials, butsuffer from severe memory access inefficiency and data transfer overhead underthe canonical one-graph-at-a-time training pattern. To tackle the challenges,we propose PiPAD, a $\underline{\textbf{Pi}}pelined$ and$\underline{\textbf{PA}}rallel$ $\underline{\textbf{D}}GNN$ training frameworkfor the end-to-end performance optimization on GPUs. From both the algorithmand runtime level, PiPAD holistically reconstructs the overall trainingparadigm from the data organization to computation manner. Capable ofprocessing multiple graph snapshots in parallel, PiPAD eliminates theunnecessary data transmission and alleviates memory access inefficiency toimprove the overall performance. Our evaluation across various datasets showsPiPAD achieves $1.22\times$-$9.57\times$ speedup over the state-of-the-art DGNNframeworks on three representative models.

PiPAD: Pipelined and Parallel Dynamic GNN Training on GPUs