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

The Strong Lottery Ticket Hypothesis (SLTH) demonstrates the existence ofhigh-performing subnetworks within a randomly initialized model, discoverablethrough pruning a convolutional neural network (CNN) without any weighttraining. A recent study, called Untrained GNNs Tickets (UGT), expanded SLTHfrom CNNs to shallow graph neural networks (GNNs). However, discrepanciespersist when comparing baseline models with learned dense weights.Additionally, there remains an unexplored area in applying SLTH to deeper GNNs,which, despite delivering improved accuracy with additional layers, suffer fromexcessive memory requirements. To address these challenges, this work utilizesMulticoated Supermasks (M-Sup), a scalar pruning mask method, and implements itin GNNs by proposing a strategy for setting its pruning thresholds adaptively.In the context of deep GNNs, this research uncovers the existence of untrainedrecurrent networks, which exhibit performance on par with their trainedfeed-forward counterparts. This paper also introduces the Multi-Stage Foldingand Unshared Masks methods to expand the search space in terms of botharchitecture and parameters. Through the evaluation of various datasets,including the Open Graph Benchmark (OGB), this work establishes a triple-winscenario for SLTH-based GNNs: by achieving high sparsity, competitiveperformance, and high memory efficiency with up to 98.7\% reduction, itdemonstrates suitability for energy-efficient graph processing.

Multicoated and Folded Graph Neural Networks with Strong Lottery Tickets