Attribute-Topology Cross-Frequency Aligned Graph Neural Networks for Homophilic and Heterophilic Graphs in Node...

Area of Science:

• Graph Neural Networks
• Machine Learning
• Data Science

Background:

• Graph neural networks (GNNs) excel at processing graph-structured data.
• Existing GNNs struggle with attribute-topology interference and overlook high-frequency graph signal information, especially in heterophilic graphs.

Purpose of the Study:

• To introduce a novel GNN architecture, Attribute-Topology Cross-Frequency Aligned (ATCFA) GNNs.
• To address the limitations of current GNNs in handling attribute-topology interference and capturing high-frequency information.

Main Methods:

• ATCFA GNNs integrate low- and high-pass filters for comprehensive topological and attribute representation.
• Frequency-specific constraints are applied to minimize noise and redundancy within each frequency band.
• Dynamic associations between topology and attribute frequency components enable interactive fusion and interference mitigation.

Main Results:

• ATCFA GNNs demonstrated superior node classification accuracy compared to state-of-the-art methods.
• The model effectively handles both homophilic and heterophilic graph structures.
• Experiments confirmed the model's ability to mitigate interference and utilize complementary information across domains.

Conclusions:

• ATCFA GNNs offer a robust solution for node classification challenges in graph representation learning.
• The proposed method effectively balances attribute and topology information across different frequency domains.
• ATCFA GNNs advance the capabilities of GNNs for complex graph data analysis.