AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Certifiable Robustness to Graph Perturbations

Aleksandar Bojchevski, Stephan Günnemann

NeurIPS 2019 · Adversarial robustness and security › Attacks on text and graphs · approach: Graph neural network robustness · 52 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Train ρ-PPNP with robust cross-entropy instead of standard cross-entropy (ablation)ρ-PPNP · Citeseer test splitF1 score: 0.70 -> 0.72Improved
Add a robust hinge-margin penalty to standard cross-entropyρ-PPNP · Citeseer test splitF1 score: 0.70 -> 0.73Improved
Use robust cross-entropy rather than the hinge-margin loss (ablation)ρ-PPNP robust training · Citeseercertifiable robustnessMixed
Robustly train with either RCE or CEM at local attack strength s=10ρ-PPNP · Citeseerpercentage of certifiably robust nodes: around 13% increaseImproved
Lower the damping factor αρ-PPNP · Cora-ML and Citeseercertifiable robustnessImproved
Lower the damping factor αρ-PPNP · Cora-ML and Citeseeraccuracy: at most 0.5%No clear effect
Restrict the attacker with a global edge-perturbation budget in addition to local budgetsρ-PPNP certificates · Cora-ML and Citeseernumber of certifiably robust nodesImproved
Allow edge additions as well as removals instead of allowing only edge removalsρ-PPNP · Cora-ML, Citeseer, and Pubmedpercentage of certifiably robust nodesHurt
Use ρ-PPNP rather than label propagationsemi-supervised node classification · Cora-ML and Citeseercertifiable robustnessImproved
Use feature propagation rather than ρ-PPNPsemi-supervised node classification · Cora-ML and Citeseercertifiable robustnessHurt
Increase local attack strength and its corresponding local perturbation budgetρ-PPNP certificates · Cora-ML and Citeseerpercentage of certifiably robust nodesHurt
Train using local attack strength s=10 rather than s=6 (ablation)ρ-PPNP robust training · Citeseercertifiable robustnessImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy