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

Self-supervised learning (SSL) for graph neural networks (GNNs) has attractedincreasing attention from the graph machine learning community in recent years,owing to its capability to learn performant node embeddings without costlylabel information. One weakness of conventional SSL frameworks for GNNs is thatthey learn through a single philosophy, such as mutual information maximizationor generative reconstruction. When applied to various downstream tasks, theseframeworks rarely perform equally well for every task, because one philosophymay not span the extensive knowledge required for all tasks. To enhance thetask generalization across tasks, as an important first step forward inexploring fundamental graph models, we introduce PARETOGNN, a multi-task SSLframework for node representation learning over graphs. Specifically, PARETOGNNis self-supervised by manifold pretext tasks observing multiple philosophies.To reconcile different philosophies, we explore a multiple-gradient descentalgorithm, such that PARETOGNN actively learns from every pretext task whileminimizing potential conflicts. We conduct comprehensive experiments over fourdownstream tasks (i.e., node classification, node clustering, link prediction,and partition prediction), and our proposal achieves the best overallperformance across tasks on 11 widely adopted benchmark datasets. Besides, weobserve that learning from multiple philosophies enhances not only the taskgeneralization but also the single task performances, demonstrating thatPARETOGNN achieves better task generalization via the disjoint yetcomplementary knowledge learned from different philosophies. Our code ispublicly available at https://github.com/jumxglhf/ParetoGNN.

Multi-task Self-supervised Graph Neural Networks Enable Stronger Task Generalization