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

Link prediction (LP) has been recognized as an important task in graphlearning with its broad practical applications. A typical application of LP isto retrieve the top scoring neighbors for a given source node, such as thefriend recommendation. These services desire the high inference scalability tofind the top scoring neighbors from many candidate nodes at low latencies.There are two popular decoders that the recent LP models mainly use to computethe edge scores from node embeddings: the HadamardMLP and Dot Product decoders.After theoretical and empirical analysis, we find that the HadamardMLP decodersare generally more effective for LP. However, HadamardMLP lacks the scalabilityfor retrieving top scoring neighbors on large graphs, since to the best of ourknowledge, there does not exist an algorithm to retrieve the top scoringneighbors for HadamardMLP decoders in sublinear complexity. To make HadamardMLPscalable, we propose the Flashlight algorithm to accelerate the top scoringneighbor retrievals for HadamardMLP: a sublinear algorithm that progressivelyapplies approximate maximum inner product search (MIPS) techniques withadaptively adjusted query embeddings. Empirical results show that Flashlightimproves the inference speed of LP by more than 100 times on the largeOGBL-CITATION2 dataset without sacrificing effectiveness. Our work paves theway for large-scale LP applications with the effective HadamardMLP decoders bygreatly accelerating their inference.

Flashlight: Scalable Link Prediction with Effective Decoders