Graph neural networks (GNNs) have been increasingly deployed in various applications that involve learning on non-Euclidean data.
Many widely used datasets for graph machine learning tasks have generally been homophilous, where nodes with similar labels connect to each other.
In this paper, we propose GARNET, a scalable spectral method to boost the adversarial robustness of GNN models for both homophilic and heterophilic graphs.
Much data with graph structures satisfy the principle of homophily, meaning that connected nodes tend to be similar with respect to a specific attribute.
Ranked #1 on Node Classification on Yelp-Fraud