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 Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Recent works show that Graph Neural Networks (GNNs) are highly non-robustwith respect to adversarial attacks on both the graph structure and the nodeattributes, making their outcomes unreliable. We propose the first method forcertifiable (non-)robustness of graph convolutional networks with respect toperturbations of the node attributes. We consider the case of binary nodeattributes (e.g. bag-of-words) and perturbations that are L_0-bounded. If anode has been certified with our method, it is guaranteed to be robust underany possible perturbation given the attack model. Likewise, we can certifynon-robustness. Finally, we propose a robust semi-supervised training procedurethat treats the labeled and unlabeled nodes jointly. As shown in ourexperimental evaluation, our method significantly improves the robustness ofthe GNN with only minimal effect on the predictive accuracy.

Certifiable Robustness and Robust Training for Graph Convolutional Networks