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

Recent works have impressively demonstrated that there exists a subnetwork inrandomly initialized convolutional neural networks (CNNs) that can match theperformance of the fully trained dense networks at initialization, without anyoptimization of the weights of the network (i.e., untrained networks). However,the presence of such untrained subnetworks in graph neural networks (GNNs)still remains mysterious. In this paper we carry out the first-of-its-kindexploration of discovering matching untrained GNNs. With sparsity as the coretool, we can find \textit{untrained sparse subnetworks} at the initialization,that can match the performance of \textit{fully trained dense} GNNs. Besidesthis already encouraging finding of comparable performance, we show that thefound untrained subnetworks can substantially mitigate the GNN over-smoothingproblem, hence becoming a powerful tool to enable deeper GNNs without bells andwhistles. We also observe that such sparse untrained subnetworks have appealingperformance in out-of-distribution detection and robustness of inputperturbations. We evaluate our method across widely-used GNN architectures onvarious popular datasets including the Open Graph Benchmark (OGB).

You Can Have Better Graph Neural Networks by Not Training Weights at All: Finding Untrained GNNs Tickets