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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

In recent years, graph neural networks (GNN) have achieved significantdevelopments in a variety of graph analytical tasks. Nevertheless, GNN'ssuperior performance will suffer from serious damage when the collected nodefeatures or structure relationships are partially missing owning to numerousunpredictable factors. Recently emerged graph completion learning (GCL) hasreceived increasing attention, which aims to reconstruct the missing nodefeatures or structure relationships under the guidance of a specificallysupervised task. Although these proposed GCL methods have made great success,they still exist the following problems: the reliance on labels, the bias ofthe reconstructed node features and structure relationships. Besides, thegeneralization ability of the existing GCL still faces a huge challenge whenboth collected node features and structure relationships are partially missingat the same time. To solve the above issues, we propose a more general GCLframework with the aid of self-supervised learning for improving the taskperformance of the existing GNN variants on graphs with features and structuremissing, termed unsupervised GCL (UGCL). Specifically, to avoid the mismatchbetween missing node features and structure during the message-passing processof GNN, we separate the feature reconstruction and structure reconstruction anddesign its personalized model in turn. Then, a dual contrastive loss on thestructure level and feature level is introduced to maximize the mutualinformation of node representations from feature reconstructing and structurereconstructing paths for providing more supervision signals. Finally, thereconstructed node features and structure can be applied to the downstream nodeclassification task. Extensive experiments on eight datasets, three GNNvariants and five missing rates demonstrate the effectiveness of our proposedmethod.

Towards Unsupervised Graph Completion Learning on Graphs with Features and Structure Missing