Signed random walk diffusion for effective representation learning in signed graphs.

Area of Science:

• Graph Neural Networks
• Social Network Analysis
• Machine Learning

Background:

• Signed social graphs model trust relationships, attracting research in representation learning.
• Existing methods like network embedding and graph convolutional networks (GCNs) have limitations.
• GCNs face performance degradation with increased depth, and network embedding is not task-specific.

Purpose of the Study:

• To propose an end-to-end graph neural network for link sign prediction in signed social graphs.
• To develop a novel random walk-based feature aggregation method tailored for signed graphs.
• To enhance node representation learning for accurate inference of missing edge signs.

Main Methods:

• Introduced Signed Diffusion Network (SidNet), a novel graph neural network architecture.
• Developed a specialized random walk-based feature aggregation for signed graphs.
• Implemented end-to-end node representation learning for link sign prediction.

Main Results:

• SidNet effectively diffuses hidden node features by leveraging information from neighboring nodes.
• Extensive experiments demonstrate SidNet's superior performance compared to state-of-the-art models.
• Achieved significant improvements in link sign prediction accuracy.

Conclusions:

• SidNet offers an effective end-to-end solution for link sign prediction in signed social graphs.
• The proposed random walk feature aggregation enhances the model's ability to capture signed graph properties.
• SidNet represents a significant advancement in analyzing trust relationships within social networks.