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 Contrastive Learning (GCL) has proven highly effective in promoting theperformance of Semi-Supervised Node Classification (SSNC). However, existingGCL methods are generally transferred from other fields like CV or NLP, whoseunderlying working mechanism remains under-explored. In this work, we firstdeeply probe the working mechanism of GCL in SSNC, and find that the promotionbrought by GCL is severely unevenly distributed: the improvement mainly comesfrom subgraphs with less annotated information, which is fundamentallydifferent from contrastive learning in other fields. However, existing GCLmethods generally ignore this uneven distribution of annotated information andapply GCL evenly to the whole graph. To remedy this issue and further improveGCL in SSNC, we propose the Topology InFormation gain-Aware Graph ContrastiveLearning (TIFA-GCL) framework that considers the annotated informationdistribution across graph in GCL. Extensive experiments on six benchmark graphdatasets, including the enormous OGB-Products graph, show that TIFA-GCL canbring a larger improvement than existing GCL methods in both transductive andinductive settings. Further experiments demonstrate the generalizability andinterpretability of TIFA-GCL.

Rethinking the Promotion Brought by Contrastive Learning to Semi-Supervised Node Classification