A graph transformer defence against graph perturbation by a flexible-pass filter

Yonghua Zhu, Jincheng Huang, Yang Chen, Robert Amor, Michael Witbrock

Abstract: Highlights•A new graph transformer is proposed to defend against graph perturbations.•It is the first Transformer with a flexible-pass filter for graph data.•Theoretical analysis reveals Transformer functions as a low-pass filter.•Contrastive learning converts the Transformer to a high-pass filter.•Empirical experiments demonstrate strong defense capabilities over GNNs.