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) mainly rely on the message-passing paradigm topropagate node features and build interactions, and different graph learningtasks require different ranges of node interactions. In this work, we explorethe capacity of GNNs to capture interactions between nodes under contexts withdifferent complexities. We discover that GNNs are usually unable to capture themost informative kinds of interaction styles for diverse graph learning tasks,and thus name this phenomenon as GNNs' representation bottleneck. As aresponse, we demonstrate that the inductive bias introduced by existing graphconstruction mechanisms can prevent GNNs from learning interactions of the mostappropriate complexity, i.e., resulting in the representation bottleneck. Toaddress that limitation, we propose a novel graph rewiring approach based oninteraction patterns learned by GNNs to adjust the receptive fields of eachnode dynamically. Extensive experiments on both real-world and syntheticdatasets prove the effectiveness of our algorithm to alleviate therepresentation bottleneck and its superiority to enhance the performance ofGNNs over state-of-the-art graph rewiring baselines.

Discovering and Explaining the Representation Bottleneck of Graph Neural Networks from Multi-order Interactions