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

Many emerging user-facing services adopt Graph Neural Networks (GNNs) toimprove serving accuracy. When the graph used by a GNN model changes,representations (embedding) of nodes in the graph should be updatedaccordingly. However, the node representation update is too slow, resulting ineither long response latency of user queries (the inference is performed afterthe update completes) or high staleness problem (the inference is performedbased on stale data). Our in-depth analysis shows that the slow update ismainly due to neighbor explosion problem in graphs and duplicated computation.Based on such findings, we propose STAG, a GNN serving framework that enableslow latency and low staleness of GNN-based services. It comprises acollaborative serving mechanism and an additivity-based incremental propagationstrategy. With the collaborative serving mechanism, only part of noderepresentations are updated during the update phase, and the finalrepresentations are calculated in the inference phase. It alleviates theneighbor explosion problem. The additivity-based incremental propagationstrategy reuses intermediate data during the update phase, eliminatingduplicated computation problem. Experimental results show that STAG acceleratesthe update phase by 1.3x~90.1x, and greatly reduces staleness time with aslight increase in response latency.

STAG: Enabling Low Latency and Low Staleness of GNN-based Services with Dynamic Graphs