GegenNet: Spectral Convolutional Neural Networks for Link Sign Prediction in Signed Bipartite Graphs

2025-08-27

AI Summary

Existing methods for link sign prediction in signed bipartite graphs (SBGs) neglect node heterogeneity and the intrinsic bipartite structure, while conventional spectral convolutions fail to capture协同 interactions between positive and negative edges. Method: We propose GegenNet, a novel spectral graph neural network leveraging Gegenbauer polynomials as learnable graph filters. It is the first to embed sign-aware mechanisms into multi-layer spectral convolution, explicitly modeling both bipartite topology and edge-sign semantics via joint inter- and intra-partition message passing. GegenNet initializes node features via fast spectral decomposition and supports alternating convolution over positive and negative edges. Results: Evaluated on six benchmark SBG datasets, GegenNet consistently outperforms 11 strong baselines, achieving up to 4.28% improvement in AUC and 11.69% in F1-score—demonstrating its superior capability in capturing SBG-specific structural and semantic properties.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Social Networks and Social Media: Influence propagation, information diffusion, and the prediction on networks

Abstract

Given a signed bipartite graph (SBG) G with two disjoint node sets U and V, the goal of link sign prediction is to predict the signs of potential links connecting U and V based on known positive and negative edges in G. The majority of existing solutions towards link sign prediction mainly focus on unipartite signed graphs, which are sub-optimal due to the neglect of node heterogeneity and unique bipartite characteristics of SBGs. To this end, recent studies adapt graph neural networks to SBGs by introducing message-passing schemes for both inter-partition (UxV) and intra-partition (UxU or VxV) node pairs. However, the fundamental spectral convolutional operators were originally designed for positive links in unsigned graphs, and thus, are not optimal for inferring missing positive or negative links from known ones in SBGs. Motivated by this, this paper proposes GegenNet, a novel and effective spectral convolutional neural network model for link sign prediction in SBGs. In particular, GegenNet achieves enhanced model capacity and high predictive accuracy through three main technical contributions: (i) fast and theoretically grounded spectral decomposition techniques for node feature initialization; (ii) a new spectral graph filter based on the Gegenbauer polynomial basis; and (iii) multi-layer sign-aware spectral convolutional networks alternating Gegenbauer polynomial filters with positive and negative edges. Our extensive empirical studies reveal that GegenNet can achieve significantly superior performance (up to a gain of 4.28% in AUC and 11.69% in F1) in link sign prediction compared to 11 strong competitors over 6 benchmark SBG datasets.

Problem

• Predicting signs of potential links in signed bipartite graphs
• Addressing limitations of existing unipartite-focused approaches
• Overcoming sub-optimal spectral convolutional operators for signed links

Innovation

• Spectral decomposition for node feature initialization
• Gegenbauer polynomial basis spectral graph filter
• Multi-layer sign-aware spectral convolutional networks