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

Few-shot node classification is tasked to provide accurate predictions fornodes from novel classes with only few representative labeled nodes. Thisproblem has drawn tremendous attention for its projection to prevailingreal-world applications, such as product categorization for newly addedcommodity categories on an E-commerce platform with scarce records or diagnosesfor rare diseases on a patient similarity graph. To tackle such challenginglabel scarcity issues in the non-Euclidean graph domain, meta-learning hasbecome a successful and predominant paradigm. More recently, inspired by thedevelopment of graph self-supervised learning, transferring pretrained nodeembeddings for few-shot node classification could be a promising alternative tometa-learning but remains unexposed. In this work, we empirically demonstratethe potential of an alternative framework, \textit{Transductive LinearProbing}, that transfers pretrained node embeddings, which are learned fromgraph contrastive learning methods. We further extend the setting of few-shotnode classification from standard fully supervised to a more realisticself-supervised setting, where meta-learning methods cannot be easily deployeddue to the shortage of supervision from training classes. Surprisingly, evenwithout any ground-truth labels, transductive linear probing withself-supervised graph contrastive pretraining can outperform thestate-of-the-art fully supervised meta-learning based methods under the sameprotocol. We hope this work can shed new light on few-shot node classificationproblems and foster future research on learning from scarcely labeled instanceson graphs.

Transductive Linear Probing: A Novel Framework for Few-Shot Node Classification