A graph transformer defence against graph perturbation by a flexible-pass filter
Yonghua Zhu, Jincheng Huang, Yang Chen, Robert Amor, Michael Witbrock · Information Fusion · 2024
Graph perturbation hinders graph models in real applications, and thus, defense methods against graph perturbation have been attracting increasing attention. However, current defense methods limit expressiveness and demand expert knowledge. To overcome these issues, in this paper, we propose a flexible-frequency graph transformer, building on the powerful expressive ability of self-attention. Specifically, we design a frequency-extraction self-attention with three heads to extract multi-frequency representations, i.e., low-frequency representation, hybrid-frequency representation, and high-frequency representation. We then design an adaptive fusion method to combine diverse representations for outputting a flexible-frequency representation. This improves the model’s expressive ability by using comprehensive information to enhance defense ability against graph perturbation. To achieve this, we adaptively capture the robust graph filters within self-attention, eliminating the need for expert knowledge. To further enhance the effectiveness of self-attention, we incorporate graph learning to capture graph information before conducting node representation learning through self-attention. Furthermore, we theoretically analyze the feasibility of our proposed method. Extensive experimental results demonstrate that our proposed method offers a dynamic and effective defense against graph perturbation compared to existing state-of-the-art methods.